Release v3.8.0

.github/workflows/update-api-md.yml

@@ -0,0 +1,57 @@
+name: Update API Reference
+
+on:
+  push:
+    branches: [ "main" ]
+    paths:
+      - 'src/praisonai-agents/praisonaiagents/**/*.py'
+      - 'src/praisonai/praisonai/**/*.py'
+      - 'src/praisonai-ts/src/**/*.ts'
+      - '.github/workflows/update-api-md.yml'
+
+permissions:
+  contents: write
+
+jobs:
+  update-api-md:
+    runs-on: ubuntu-latest
+    
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+      with:
+        ref: main
+        persist-credentials: true
+        fetch-depth: 0
+
+    - name: Set up Python
+      uses: actions/setup-python@v5
+      with:
+        python-version: '3.11'
+
+    - name: Generate api.md
+      run: |
+        cd $GITHUB_WORKSPACE
+        python3 src/praisonai/scripts/generate_api_md.py
+        
+    - name: Check for changes
+      id: check_changes
+      run: |
+        if git diff --quiet api.md; then
+          echo "changes=false" >> $GITHUB_OUTPUT
+          echo "No changes to api.md"
+        else
+          echo "changes=true" >> $GITHUB_OUTPUT
+          echo "api.md has been updated"
+          git diff api.md | head -50
+        fi
+
+    - name: Commit and push changes
+      if: steps.check_changes.outputs.changes == 'true'
+      uses: stefanzweifel/git-auto-commit-action@v5
+      with:
+        commit_message: "docs: auto-update api.md [skip ci]"
+        file_pattern: 'api.md'
+        commit_user_name: "github-actions[bot]"
+        commit_user_email: "41898282+github-actions[bot]@users.noreply.github.com"
+        commit_author: "github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>"

.gitignore

@@ -83,4 +83,5 @@ issue_details.json
 # MCP Registry tokens
 .mcpregistry_*
 
-*.log
+*.log
+.internal

api.md

@@ -552,7 +552,7 @@ Methods:
 
 Types:
 ```python
-from praisonaiagents import AutoRagConfig, Citation, CodeExecutionStep, DeepResearchResponse, ExpandResult, ExpandStrategy, FileSearchCall, MCPCall, Provider, RagRetrievalPolicy, ReasoningStep, RewriteResult, RewriteStrategy, Task, WebSearchCall
+from praisonaiagents import AutoRagConfig, Citation, CodeExecutionStep, DeepResearchResponse, ExpandResult, ExpandStrategy, FileSearchCall, MCPCall, Provider, RagRetrievalPolicy, ReasoningStep, RewriteResult, RewriteStrategy, Task, WebSearchCall, warmup
 ```
 
 Methods:
@@ -569,6 +569,7 @@ Methods:
 * <code title="class Task">Task.<a href="./src/praisonai-agents/praisonaiagents/task/task.py">execute_callback_sync</a>(task_output: TaskOutput) -> None</code>
 * <code title="class Task">Task.<a href="./src/praisonai-agents/praisonaiagents/task/task.py">initialize_memory</a>()</code>
 * <code title="class Task">Task.<a href="./src/praisonai-agents/praisonaiagents/task/task.py">store_in_memory</a>(content: str, agent_name: str = None, task_id: str = None)</code>
+* <code title="function">praisonaiagents.<a href="./src/praisonai-agents/praisonaiagents/__init__.py">warmup</a>(include_litellm: bool = False, include_openai: bool = True) -> dict</code>
 
 # Wrapper (praisonai)
 
@@ -731,6 +732,10 @@ Methods:
 * <code title="cli">praisonai skills install <a href="./src/praisonai/praisonai/cli/commands/skills.py">--help</a></code>
 * <code title="cli">praisonai skills list <a href="./src/praisonai/praisonai/cli/commands/skills.py">--help</a></code>
 * <code title="cli">praisonai skills validate <a href="./src/praisonai/praisonai/cli/commands/skills.py">--help</a></code>
+* <code title="cli">praisonai standardise check <a href="./src/praisonai/praisonai/cli/commands/standardise.py">--help</a></code>
+* <code title="cli">praisonai standardise fix <a href="./src/praisonai/praisonai/cli/commands/standardise.py">--help</a></code>
+* <code title="cli">praisonai standardise init <a href="./src/praisonai/praisonai/cli/commands/standardise.py">--help</a></code>
+* <code title="cli">praisonai standardise report <a href="./src/praisonai/praisonai/cli/commands/standardise.py">--help</a></code>
 * <code title="cli">praisonai templates create <a href="./src/praisonai/praisonai/cli/commands/templates.py">--help</a></code>
 * <code title="cli">praisonai templates list <a href="./src/praisonai/praisonai/cli/commands/templates.py">--help</a></code>
 * <code title="cli">praisonai test info <a href="./src/praisonai/praisonai/cli/commands/test.py">--help</a></code>

docker/Dockerfile.chat

@@ -16,7 +16,7 @@ RUN mkdir -p /root/.praison
 # Install Python packages (using latest versions)
 RUN pip install --no-cache-dir \
     praisonai_tools \
-    "praisonai>=3.7.3" \
+    "praisonai>=3.8.0" \
     "praisonai[chat]" \
     "embedchain[github,youtube]"
 

docker/Dockerfile.dev

@@ -20,7 +20,7 @@ RUN mkdir -p /root/.praison
 # Install Python packages (using latest versions)
 RUN pip install --no-cache-dir \
     praisonai_tools \
-    "praisonai>=3.7.3" \
+    "praisonai>=3.8.0" \
     "praisonai[ui]" \
     "praisonai[chat]" \
     "praisonai[realtime]" \

docker/Dockerfile.ui

@@ -16,7 +16,7 @@ RUN mkdir -p /root/.praison
 # Install Python packages (using latest versions)
 RUN pip install --no-cache-dir \
     praisonai_tools \
-    "praisonai>=3.7.3" \
+    "praisonai>=3.8.0" \
     "praisonai[ui]" \
     "praisonai[crewai]"
 

src/praisonai-agents/praisonaiagents/agent/agent.py

@@ -44,7 +44,7 @@ _shared_apps = {}  # Dict of port -> FastAPI app
 if TYPE_CHECKING:
     from ..task.task import Task
     from ..main import TaskOutput
-    from ..handoff import Handoff
+    from .handoff import Handoff, HandoffConfig, HandoffResult
     from ..rag.models import RAGResult, ContextPack
 
 class Agent:
@@ -291,96 +291,87 @@ class Agent:
         """Initialize an Agent instance.
 
         Args:
-            name (Optional[str], optional): Name of the agent used for identification and logging.
-                If None, defaults to "Agent". Defaults to None.
-            role (Optional[str], optional): Role or job title that defines the agent's expertise
-                and behavior patterns. Examples: "Data Analyst", "Content Writer". Defaults to None.
-            goal (Optional[str], optional): Primary objective or goal the agent aims to achieve.
-                Defines the agent's purpose and success criteria. Defaults to None.
-            backstory (Optional[str], optional): Background story or context that shapes the agent's
-                personality and decision-making approach. Defaults to None.
-            instructions (Optional[str], optional): Direct instructions that override role, goal,
-                and backstory when provided. Used for simple, task-specific agents. Defaults to None.
-            llm (Optional[Union[str, Any]], optional): Language model configuration. Can be a model
-                name string (e.g., "gpt-5-nano", "anthropic/claude-3-sonnet") or a configured LLM object.
-                Defaults to environment variable OPENAI_MODEL_NAME or "gpt-5-nano".
-            tools (Optional[List[Any]], optional): List of tools, functions, or capabilities
-                available to the agent for task execution. Can include callables, tool objects,
-                or MCP instances. Defaults to None.
-            function_calling_llm (Optional[Any], optional): Dedicated language model for function
-                calling operations. If None, uses the main llm parameter. Defaults to None.
-            max_iter (int, optional): Maximum number of iterations the agent can perform during
-                task execution to prevent infinite loops. Defaults to 20.
-            max_rpm (Optional[int], optional): Maximum requests per minute to rate limit API calls
-                and prevent quota exhaustion. If None, no rate limiting is applied. Defaults to None.
-            max_execution_time (Optional[int], optional): Maximum execution time in seconds for
-                agent operations before timeout. If None, no time limit is enforced. Defaults to None.
-            memory (Optional[Any], optional): Memory system for storing and retrieving information
-                across conversations. Requires memory dependencies to be installed. Defaults to None.
-            verbose (bool, optional): Enable detailed logging and status updates during agent
-                execution for debugging and monitoring. Defaults to True.
-            allow_delegation (bool, optional): Allow the agent to delegate tasks to other agents
-                or sub-processes when appropriate. Defaults to False.
-            step_callback (Optional[Any], optional): Callback function called after each step
-                of agent execution for custom monitoring or intervention. Defaults to None.
-            cache (bool, optional): Enable caching of responses and computations to improve
-                performance and reduce API costs. Defaults to True.
-            system_template (Optional[str], optional): Custom template for system prompts that
-                overrides the default system prompt generation. Defaults to None.
-            prompt_template (Optional[str], optional): Template for formatting user prompts
-                before sending to the language model. Defaults to None.
-            response_template (Optional[str], optional): Template for formatting agent responses
-                before returning to the user. Defaults to None.
-            allow_code_execution (Optional[bool], optional): Enable the agent to execute code
-                snippets during task completion. Use with caution for security. Defaults to False.
-            max_retry_limit (int, optional): Maximum number of retry attempts for failed operations
-                before giving up. Helps handle transient errors. Defaults to 2.
-            code_execution_mode (Literal["safe", "unsafe"], optional): Safety mode for code execution.
-                "safe" restricts dangerous operations, "unsafe" allows full code execution. Defaults to "safe".
-            embedder_config (Optional[Dict[str, Any]], optional): Configuration dictionary for
-                text embedding models used in knowledge retrieval and similarity search. Defaults to None.
-            knowledge (Optional[List[str]], optional): List of knowledge sources (file paths, URLs,
-                or text content) to be processed and made available to the agent. Defaults to None.
-            knowledge_config (Optional[Dict[str, Any]], optional): Configuration for knowledge
-                processing and retrieval system including chunking and indexing parameters. Defaults to None.
-            use_system_prompt (Optional[bool], optional): Whether to include system prompts in
-                conversations to establish agent behavior and context. Defaults to True.
-            markdown (bool, optional): Enable markdown formatting in agent responses for better
-                readability and structure. Defaults to True.
-            stream (bool, optional): Enable streaming responses from the language model for real-time
-                output when using Agent.start() method. Defaults to False for backward compatibility.
-            metrics (bool, optional): Enable automatic token usage tracking and display summary
-                when tasks complete. Simplifies token monitoring for cost optimization. Defaults to False.
-            self_reflect (bool, optional): Enable self-reflection capabilities where the agent
-                evaluates and improves its own responses. Defaults to False.
-            max_reflect (int, optional): Maximum number of self-reflection iterations to prevent
-                excessive reflection loops. Defaults to 3.
-            min_reflect (int, optional): Minimum number of self-reflection iterations required
-                before accepting a response as satisfactory. Defaults to 1.
-            reflect_llm (Optional[str], optional): Dedicated language model for self-reflection
-                operations. If None, uses the main llm parameter. Defaults to None.
-            reflect_prompt (Optional[str], optional): Custom prompt template for self-reflection
-                that guides the agent's self-evaluation process. Defaults to None.
-            user_id (Optional[str], optional): Unique identifier for the user or session to
-                enable personalized responses and memory isolation. Defaults to "praison".
-            reasoning_steps (bool, optional): Enable step-by-step reasoning output to show the
-                agent's thought process during problem solving. Defaults to False.
-            guardrail (Optional[Union[Callable[['TaskOutput'], Tuple[bool, Any]], str]], optional):
-                Safety mechanism to validate agent outputs. Can be a validation function or
-                description string for LLM-based validation. Defaults to None.
-            max_guardrail_retries (int, optional): Maximum number of retry attempts when guardrail
-                validation fails before giving up. Defaults to 3.
-            handoffs (Optional[List[Union['Agent', 'Handoff']]], optional): List of agents or
-                handoff configurations that this agent can delegate tasks to. Enables agent-to-agent
-                collaboration and task specialization. Defaults to None.
-            base_url (Optional[str], optional): Base URL for custom LLM endpoints (e.g., Ollama).
-                If provided, automatically creates a custom LLM instance. Defaults to None.
-            api_key (Optional[str], optional): API key for LLM provider. If not provided,
-                falls back to environment variables. Defaults to None.
+            name: Agent name for identification and logging. Defaults to "Agent".
+            role: Role/job title defining expertise (e.g., "Data Analyst").
+            goal: Primary objective the agent aims to achieve.
+            backstory: Background context shaping personality and decisions.
+            instructions: Direct instructions (overrides role/goal/backstory). Recommended for simple agents.
+            llm: Model name string ("gpt-4o", "anthropic/claude-3-sonnet") or LLM object.
+                Defaults to OPENAI_MODEL_NAME env var or "gpt-5-nano".
+            model: Alias for llm parameter.
+            function_calling_llm: Dedicated LLM for function calling. Deprecated: use llm=.
+            llm_config: LLM configuration dict. Deprecated: use llm=.
+            base_url: Custom LLM endpoint URL (e.g., for Ollama). Kept separate for auth.
+            api_key: API key for LLM provider. Kept separate for auth.
+            tools: List of tools, functions, callables, or MCP instances.
+            allow_delegation: Allow task delegation to other agents. Defaults to False.
+            allow_code_execution: Enable code execution during tasks. Defaults to False.
+            code_execution_mode: "safe" (restricted) or "unsafe" (full access). Defaults to "safe".
+            handoffs: List of Agent or Handoff objects for agent-to-agent collaboration.
+            auto_save: Session name for automatic session saving.
+            rate_limiter: Rate limiter instance for API call throttling.
+            memory: Memory system configuration. Accepts:
+                - bool: True enables defaults, False disables
+                - MemoryConfig: Custom configuration
+                - MemoryManager: Pre-configured instance
+            knowledge: Knowledge sources. Accepts:
+                - bool: True enables defaults
+                - List[str]: File paths, URLs, or text content
+                - KnowledgeConfig: Custom configuration
+            planning: Planning mode. Accepts:
+                - bool: True enables with defaults
+                - PlanningConfig: Custom configuration
+            reflection: Self-reflection. Accepts:
+                - bool: True enables with defaults
+                - ReflectionConfig: Custom configuration
+            guardrails: Output validation. Accepts:
+                - bool: True enables with defaults
+                - Callable: Validation function
+                - GuardrailConfig: Custom configuration
+            web: Web search/fetch. Accepts:
+                - bool: True enables with defaults
+                - WebConfig: Custom configuration
+            context: Context management. Accepts:
+                - bool: True enables with defaults
+                - ManagerConfig: Custom configuration
+            autonomy: Autonomy settings. Accepts:
+                - bool: True enables with defaults
+                - Dict: Configuration dict
+                - AutonomyConfig: Custom configuration
+            verification_hooks: List of VerificationHook instances for output verification.
+            output: Output configuration. Accepts:
+                - str: Preset name ("silent", "actions", "verbose", "json", "stream")
+                - OutputConfig: Custom configuration
+                Controls: verbose, markdown, stream, metrics, reasoning_steps
+            execution: Execution configuration. Accepts:
+                - str: Preset name ("fast", "balanced", "thorough")
+                - ExecutionConfig: Custom configuration
+                Controls: max_iter, max_rpm, max_execution_time, max_retry_limit
+            templates: Template configuration (TemplateConfig).
+                Controls: system_template, prompt_template, response_template
+            caching: Caching configuration. Accepts:
+                - bool: True enables with defaults
+                - CachingConfig: Custom configuration
+            hooks: Event hooks. Accepts:
+                - List: List of hook callables
+                - HooksConfig: Custom configuration
+            skills: Agent skills. Accepts:
+                - List[str]: Skill directory paths
+                - SkillsConfig: Custom configuration
 
         Raises:
             ValueError: If all of name, role, goal, backstory, and instructions are None.
             ImportError: If memory or LLM features are requested but dependencies are not installed.
+
+        Note:
+            Many legacy parameters have been consolidated into config objects:
+            - verbose, markdown, stream, metrics, reasoning_steps → output=
+            - max_iter, max_rpm, max_execution_time, max_retry_limit → execution=
+            - self_reflect, max_reflect, min_reflect, reflect_llm → reflection=
+            - guardrail, max_guardrail_retries → guardrails=
+            - system_template, prompt_template, response_template → templates=
+            - cache, prompt_caching → caching=
+            - web_search, web_fetch → web=
         """
         # Add check at start if memory is requested
         if memory is not None:
@@ -447,14 +438,27 @@ class Agent:
         # ─────────────────────────────────────────────────────────────────────
         # Resolve OUTPUT param using unified resolver
         # Supports: None, str preset, list [preset, overrides], OutputConfig
+        # DEFAULT: "silent" mode (zero overhead, fastest performance)
+        # Can be overridden by PRAISONAI_OUTPUT env var
         # ─────────────────────────────────────────────────────────────────────
+        from ..config.presets import DEFAULT_OUTPUT_MODE
+        
+        # Check for env var override
+        env_output = os.environ.get('PRAISONAI_OUTPUT', '').lower()
+        if env_output and env_output in OUTPUT_PRESETS and output is None:
+            output = env_output
+        
+        # Use default output mode if not specified
+        if output is None:
+            output = DEFAULT_OUTPUT_MODE
+        
         _output_config = resolve(
             value=output,
             param_name="output",
             config_class=OutputConfig,
             presets=OUTPUT_PRESETS,
             array_mode=ArrayMode.PRESET_OVERRIDE,
-            default=OutputConfig(),
+            default=OutputConfig(),  # OutputConfig defaults to silent mode (zero overhead)
         )
         if _output_config:
             verbose = _output_config.verbose
@@ -463,8 +467,24 @@ class Agent:
             metrics = _output_config.metrics
             reasoning_steps = _output_config.reasoning_steps
             output_style = getattr(_output_config, 'style', None)
+            actions_trace = getattr(_output_config, 'actions_trace', False)  # Default False (silent)
+            json_output = getattr(_output_config, 'json_output', False)
         else:
-            verbose, markdown, stream, metrics, reasoning_steps = True, True, False, False, False
+            # Fallback defaults match silent mode (zero overhead)
+            verbose, markdown, stream, metrics, reasoning_steps = False, False, False, False, False
+            actions_trace = False  # No callbacks by default
+            json_output = False
+        
+        # Enable actions output mode if configured
+        # This registers callbacks to capture tool calls and final output
+        if actions_trace:
+            try:
+                from ..output.actions import enable_actions_mode, is_actions_mode_enabled
+                if not is_actions_mode_enabled():
+                    output_format = "jsonl" if json_output else "text"
+                    enable_actions_mode(redact=True, use_color=True, format=output_format)
+            except ImportError:
+                pass  # Actions module not available
         
         # ─────────────────────────────────────────────────────────────────────
         # Resolve EXECUTION param using unified resolver
@@ -1065,6 +1085,9 @@ Your Goal: {self.goal}
         self._context_param = context  # Store raw param for lazy init
         self._context_manager = None  # Lazy initialized on first use
         self._context_manager_initialized = False
+        
+        # Action trace mode - handled via display callbacks, not separate emitter
+        self._actions_trace = actions_trace
 
     @property
     def auto_memory(self):
@@ -4235,16 +4258,86 @@ Output MUST be JSON with 'reflection' and 'satisfactory'.
             return None
 
     async def arun(self, prompt: str, **kwargs):
-        """Async alias for astart() method"""
-        return await self.astart(prompt, **kwargs)
+        """Async version of run() - silent, non-streaming, returns structured result.
+        
+        Production-friendly async execution. Does not stream or display output.
+        
+        Args:
+            prompt: The input prompt to process
+            **kwargs: Additional arguments passed to achat()
+            
+        Returns:
+            The agent's response as a string
+        """
+        # Force non-streaming, non-display for production use
+        kwargs['stream'] = False
+        return await self.achat(prompt, **kwargs)
 
     async def astart(self, prompt: str, **kwargs):
-        """Async version of start method"""
+        """Async version of start() - interactive, streaming-aware.
+        
+        Beginner-friendly async execution. Streams by default when in TTY.
+        
+        Args:
+            prompt: The input prompt to process
+            **kwargs: Additional arguments passed to achat()
+            
+        Returns:
+            The agent's response as a string
+        """
+        import sys
+        
+        # Determine streaming behavior (same logic as start())
+        stream_requested = kwargs.get('stream')
+        if stream_requested is None:
+            if getattr(self, 'stream', None) is not None:
+                stream_requested = self.stream
+            else:
+                stream_requested = sys.stdout.isatty()
+        
+        kwargs['stream'] = stream_requested
         return await self.achat(prompt, **kwargs)
 
     def run(self, prompt: str, **kwargs):
-        """Alias for start() method"""
-        return self.start(prompt, **kwargs)
+        """Execute agent silently and return structured result.
+        
+        Production-friendly execution. Does not stream or display output by default.
+        Use this for programmatic/scripted usage where you want the result only.
+        
+        Args:
+            prompt: The input prompt to process
+            **kwargs: Additional arguments:
+                - stream (bool): Force streaming if True. Default: False
+                - display (bool): Force terminal display if True. Default: False
+                - output (str): Output preset override
+                
+        Returns:
+            The agent's response as a string
+            
+        Example:
+            ```python
+            agent = Agent(instructions="You are helpful")
+            result = agent.run("What is 2+2?")  # Silent, returns "4"
+            print(result)
+            ```
+        """
+        # Production defaults: no streaming, no display
+        if 'stream' not in kwargs:
+            kwargs['stream'] = False
+        
+        # Load history context
+        self._load_history_context()
+        
+        # Check if planning mode is enabled
+        if self.planning:
+            result = self._start_with_planning(prompt, **kwargs)
+        else:
+            result = self.chat(prompt, **kwargs)
+        
+        # Auto-save session if enabled
+        self._auto_save_session()
+        
+        return result
     
     def _get_planning_agent(self):
         """Lazy load PlanningAgent for planning mode."""
@@ -4379,18 +4472,59 @@ Write the complete compiled report:"""
         # Chat history is preserved in self.chat_history (no action needed)
 
     def start(self, prompt: str, **kwargs):
-        """Start the agent with a prompt. This is a convenience method that wraps chat()."""
-        # Load history from past sessions (now handled via context= param)
+        """Start the agent interactively with streaming output.
+        
+        Beginner-friendly execution. Streams output by default when running in a TTY.
+        Use this for interactive/terminal usage where you want to see output in real-time.
+        
+        Args:
+            prompt: The input prompt to process
+            **kwargs: Additional arguments:
+                - stream (bool | None): Override streaming. None = auto-detect TTY
+                - display (bool | None): Override display. None = auto (True if streaming)
+                - output (str): Output preset override
+                
+        Returns:
+            - If streaming: Generator yielding response chunks
+            - If not streaming: The complete response as a string
+            
+        Example:
+            ```python
+            agent = Agent(instructions="You are helpful")
+            
+            # Interactive use - streams by default in terminal
+            for chunk in agent.start("Tell me a story"):
+                print(chunk, end="", flush=True)
+            
+            # Or let it handle display automatically
+            result = agent.start("What is 2+2?")  # Streams if TTY
+            ```
+        """
+        import sys
+        
+        # Load history from past sessions
         self._load_history_context()
         
+        # Determine streaming behavior
+        # Priority: explicit kwarg > agent's stream attribute > TTY detection
+        stream_requested = kwargs.get('stream')
+        if stream_requested is None:
+            # Check agent's stream attribute first
+            if getattr(self, 'stream', None) is not None:
+                stream_requested = self.stream
+            else:
+                # Auto-detect: stream if stdout is a TTY (interactive terminal)
+                stream_requested = sys.stdout.isatty()
+        
         # Check if planning mode is enabled
         if self.planning:
             result = self._start_with_planning(prompt, **kwargs)
-        elif kwargs.get('stream', getattr(self, 'stream', False)):
+        elif stream_requested:
             # Return a generator for streaming response
+            kwargs['stream'] = True
             result = self._start_stream(prompt, **kwargs)
         else:
-            # Return regular chat response for backward compatibility
+            # Return regular chat response
             kwargs['stream'] = False
             result = self.chat(prompt, **kwargs)
         
@@ -4399,6 +4533,48 @@ Write the complete compiled report:"""
         
         return result
     
+    def iter_stream(self, prompt: str, **kwargs):
+        """Stream agent response as an iterator of chunks.
+        
+        App-friendly streaming. Yields response chunks without terminal display.
+        Use this for building custom UIs or processing streams programmatically.
+        
+        Args:
+            prompt: The input prompt to process
+            **kwargs: Additional arguments:
+                - display (bool): Show terminal output. Default: False
+                - output (str): Output preset override
+                
+        Yields:
+            str: Response chunks as they are generated
+            
+        Example:
+            ```python
+            agent = Agent(instructions="You are helpful")
+            
+            # Process stream programmatically
+            full_response = ""
+            for chunk in agent.iter_stream("Tell me a story"):
+                full_response += chunk
+                # Custom processing here
+            
+            # Or collect all at once
+            response = "".join(agent.iter_stream("Hello"))
+            ```
+        """
+        # Load history context
+        self._load_history_context()
+        
+        # Force streaming, no display by default (app-friendly)
+        kwargs['stream'] = True
+        
+        # Use the internal streaming generator
+        for chunk in self._start_stream(prompt, **kwargs):
+            yield chunk
+        
+        # Auto-save session if enabled
+        self._auto_save_session()
+    
     def _load_history_context(self):
         """Load history from past sessions into context.
         

src/praisonai-agents/praisonaiagents/config/feature_configs.py

@@ -391,10 +391,21 @@ class OutputConfig:
     
     Consolidates: verbose, markdown, stream, metrics, reasoning_steps, output_style
     
+    DEFAULT: output="silent" (zero overhead, fastest performance)
+    
     Usage:
-        # Simple preset
+        # Default is silent mode (no output overhead, programmatic use)
+        Agent(instructions="...")  # Uses output="silent"
+        
+        # Actions mode (tool calls + final output trace)
+        Agent(output="actions")
+        
+        # Verbose mode with Rich panels
         Agent(output="verbose")
         
+        # JSON mode for piping
+        Agent(output="json")
+        
         # With config
         Agent(output=OutputConfig(
             verbose=True,
@@ -404,11 +415,11 @@ class OutputConfig:
             reasoning_steps=True,
         ))
     """
-    # Verbosity
-    verbose: bool = True
+    # Verbosity - False by default for silent mode (fastest)
+    verbose: bool = False
     
-    # Formatting
-    markdown: bool = True
+    # Formatting - False by default for silent mode
+    markdown: bool = False
     
     # Streaming
     stream: bool = False
@@ -422,6 +433,14 @@ class OutputConfig:
     # Output style (custom styling)
     style: Optional[Any] = None
     
+    # Actions trace mode - shows tool calls, agent lifecycle, final output
+    # When True, registers callbacks and outputs to stderr
+    # DEFAULT: False for zero overhead (silent mode)
+    actions_trace: bool = False
+    
+    # JSON output mode - emit JSONL events for piping
+    json_output: bool = False
+    
     def to_dict(self) -> Dict[str, Any]:
         """Convert to dictionary."""
         return {
@@ -430,6 +449,8 @@ class OutputConfig:
             "stream": self.stream,
             "metrics": self.metrics,
             "reasoning_steps": self.reasoning_steps,
+            "actions_trace": self.actions_trace,
+            "json_output": self.json_output,
         }
 
 

src/praisonai-agents/praisonaiagents/config/presets.py

@@ -41,6 +41,37 @@ MEMORY_URL_SCHEMES: Dict[str, str] = {
 # =============================================================================
 
 OUTPUT_PRESETS: Dict[str, Dict[str, Any]] = {
+    # Silent preset - DEFAULT for SDK
+    # Zero overhead, no callbacks, no I/O on hot path
+    # Fastest performance for programmatic use
+    "silent": {
+        "verbose": False,
+        "markdown": False,
+        "stream": False,
+        "metrics": False,
+        "reasoning_steps": False,
+        "actions_trace": False,
+    },
+    # Actions preset - opt-in observability
+    # Shows action trace (tool calls, agent lifecycle) + final output
+    # Registers callbacks, outputs to stderr
+    "actions": {
+        "verbose": False,
+        "markdown": False,
+        "stream": False,
+        "metrics": False,
+        "reasoning_steps": False,
+        "actions_trace": True,  # Special flag for action trace mode
+    },
+    # Plain preset - final output only, no action trace
+    "plain": {
+        "verbose": False,
+        "markdown": False,
+        "stream": False,
+        "metrics": False,
+        "reasoning_steps": False,
+        "actions_trace": False,
+    },
     "minimal": {
         "verbose": False,
         "markdown": False,
@@ -69,13 +100,6 @@ OUTPUT_PRESETS: Dict[str, Dict[str, Any]] = {
         "metrics": True,
         "reasoning_steps": True,
     },
-    "silent": {
-        "verbose": False,
-        "markdown": False,
-        "stream": False,
-        "metrics": False,
-        "reasoning_steps": False,
-    },
     # Streaming preset - enables streaming by default
     "stream": {
         "verbose": True,
@@ -84,8 +108,23 @@ OUTPUT_PRESETS: Dict[str, Dict[str, Any]] = {
         "metrics": False,
         "reasoning_steps": False,
     },
+    # JSON preset - JSONL output for piping
+    "json": {
+        "verbose": False,
+        "markdown": False,
+        "stream": False,
+        "metrics": False,
+        "reasoning_steps": False,
+        "actions_trace": True,
+        "json_output": True,  # Output as JSONL
+    },
 }
 
+# Default output mode - can be overridden by PRAISONAI_OUTPUT env var
+# "silent" = zero overhead, fastest performance (DEFAULT)
+# "actions" = tool call trace + final output (opt-in observability)
+DEFAULT_OUTPUT_MODE = "silent"
+
 
 # =============================================================================
 # Execution Presets

src/praisonai-agents/praisonaiagents/llm/openai_client.py

@@ -649,6 +649,7 @@ class OpenAIClient:
             chunks = []
             first_token_emitted = False
             last_content_time = None
+            ttft_logged = False
             
             # If display function provided, use Live display
             if display_fn:
@@ -669,6 +670,12 @@ class OpenAIClient:
                             full_response_text += content
                             last_content_time = time.perf_counter()
                             
+                            # Log TTFT on first token
+                            if not ttft_logged:
+                                ttft_ms = (last_content_time - request_start_perf) * 1000
+                                logging.debug(f"TTFT (Time To First Token): {ttft_ms:.0f}ms")
+                                ttft_logged = True
+                            
                             # Emit FIRST_TOKEN on first content
                             if _emit and not first_token_emitted:
                                 stream_callback(StreamEvent(
@@ -715,11 +722,17 @@ class OpenAIClient:
                 for chunk in response_stream:
                     chunks.append(chunk)
                     
-                    if _emit and chunk.choices and chunk.choices[0].delta.content:
+                    if chunk.choices and chunk.choices[0].delta.content:
                         content = chunk.choices[0].delta.content
                         last_content_time = time.perf_counter()
                         
-                        if not first_token_emitted:
+                        # Log TTFT on first token
+                        if not ttft_logged:
+                            ttft_ms = (last_content_time - request_start_perf) * 1000
+                            logging.debug(f"TTFT (Time To First Token): {ttft_ms:.0f}ms")
+                            ttft_logged = True
+                        
+                        if _emit and not first_token_emitted:
                             stream_callback(StreamEvent(
                                 type=StreamEventType.FIRST_TOKEN,
                                 timestamp=last_content_time,
@@ -1130,7 +1143,9 @@ class OpenAIClient:
                 if display_fn and console:
                     # When verbose (display_fn provided), use streaming for better UX
                     try:
-                        with Live(display_fn("", start_time), console=console, refresh_per_second=4, transient=True) as live:
+                        request_start_perf = time.perf_counter()
+                        ttft_logged = False
+                        with Live(display_fn("", start_time), console=console, refresh_per_second=10, transient=True) as live:
                             # Use streaming when display_fn is provided for progressive display
                             response_stream = self.create_completion(
                                 messages=messages,
@@ -1148,6 +1163,11 @@ class OpenAIClient:
                             for chunk in response_stream:
                                 chunks.append(chunk)
                                 if chunk.choices[0].delta.content:
+                                    # Log TTFT on first token
+                                    if not ttft_logged:
+                                        ttft_ms = (time.perf_counter() - request_start_perf) * 1000
+                                        logging.debug(f"TTFT (Time To First Token): {ttft_ms:.0f}ms")
+                                        ttft_logged = True
                                     full_response_text += chunk.choices[0].delta.content
                                     live.update(display_fn(full_response_text, start_time))
                             

src/praisonai-agents/praisonaiagents/trace/__init__.py

@@ -0,0 +1,68 @@
+"""
+Trace Module for PraisonAI Agents.
+
+Provides lightweight action tracing for the `output="actions"` mode.
+Shows agent lifecycle events, tool calls, and final output without
+verbose internal details.
+
+Zero Performance Impact:
+- All imports are lazy loaded via __getattr__
+- NoOpSink is the default (zero overhead when not used)
+- Disabled emitter has near-zero overhead
+
+Usage:
+    from praisonaiagents import Agent
+    
+    # Simple usage - shows action trace
+    agent = Agent(instructions="...", output="actions")
+    agent.start("Do something")
+    
+    # Advanced - capture to file
+    from praisonaiagents.trace import ActionTraceConfig
+    
+    agent = Agent(
+        instructions="...",
+        output="actions",
+        trace=ActionTraceConfig(
+            sink_type="jsonl",
+            file_path="trace.jsonl",
+        )
+    )
+"""
+
+__all__ = [
+    # Core types
+    "ActionEvent",
+    "ActionEventType",
+    "ActionTraceConfig",
+    # Sinks
+    "TraceSink",
+    "NoOpSink",
+    "ListSink",
+    # Emitter
+    "TraceEmitter",
+    # Redaction
+    "redact_dict",
+    "REDACT_KEYS",
+]
+
+
+def __getattr__(name: str):
+    """Lazy load module components to avoid import overhead."""
+    if name in ("ActionEvent", "ActionEventType", "ActionTraceConfig"):
+        from .protocol import ActionEvent, ActionEventType, ActionTraceConfig
+        return locals()[name]
+    
+    if name in ("TraceSink", "NoOpSink", "ListSink"):
+        from .protocol import TraceSink, NoOpSink, ListSink
+        return locals()[name]
+    
+    if name == "TraceEmitter":
+        from .protocol import TraceEmitter
+        return TraceEmitter
+    
+    if name in ("redact_dict", "REDACT_KEYS"):
+        from .redact import redact_dict, REDACT_KEYS
+        return locals()[name]
+    
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")

src/praisonai-agents/praisonaiagents/trace/protocol.py

@@ -0,0 +1,398 @@
+"""
+Action Trace Protocol for PraisonAI Agents.
+
+Lightweight protocol for tracing agent actions. Uses only stdlib
+to ensure zero dependencies and minimal overhead.
+
+Schema Version: 1.0
+"""
+
+import json
+import time
+from dataclasses import dataclass, field
+from enum import Enum
+from typing import Any, Dict, List, Optional, Protocol, runtime_checkable
+
+
+# Schema version for forward compatibility
+SCHEMA_VERSION = "1.0"
+
+
+class ActionEventType(str, Enum):
+    """Types of action events."""
+    AGENT_START = "agent_start"
+    AGENT_END = "agent_end"
+    TOOL_START = "tool_start"
+    TOOL_END = "tool_end"
+    ERROR = "error"
+    OUTPUT = "output"
+
+
+@dataclass
+class ActionEvent:
+    """
+    Lightweight action trace event.
+    
+    Represents a single action in the agent execution trace.
+    Designed for minimal memory footprint and fast serialization.
+    """
+    event_type: str
+    timestamp: float
+    schema_version: str = SCHEMA_VERSION
+    agent_id: Optional[str] = None
+    agent_name: Optional[str] = None
+    tool_name: Optional[str] = None
+    tool_args: Optional[Dict[str, Any]] = None
+    tool_result_summary: Optional[str] = None
+    duration_ms: Optional[float] = None
+    status: Optional[str] = None  # ok, error
+    error_message: Optional[str] = None
+    metadata: Dict[str, Any] = field(default_factory=dict)
+    
+    def to_dict(self) -> Dict[str, Any]:
+        """Convert to dictionary, excluding None values."""
+        result = {
+            "schema_version": self.schema_version,
+            "event_type": self.event_type,
+            "timestamp": self.timestamp,
+        }
+        
+        # Add optional fields only if set
+        if self.agent_id is not None:
+            result["agent_id"] = self.agent_id
+        if self.agent_name is not None:
+            result["agent_name"] = self.agent_name
+        if self.tool_name is not None:
+            result["tool_name"] = self.tool_name
+        if self.tool_args is not None:
+            result["tool_args"] = self.tool_args
+        if self.tool_result_summary is not None:
+            result["tool_result_summary"] = self.tool_result_summary
+        if self.duration_ms is not None:
+            result["duration_ms"] = self.duration_ms
+        if self.status is not None:
+            result["status"] = self.status
+        if self.error_message is not None:
+            result["error_message"] = self.error_message
+        if self.metadata:
+            result["metadata"] = self.metadata
+        
+        return result
+    
+    def to_json(self) -> str:
+        """Convert to JSON string."""
+        return json.dumps(self.to_dict(), default=str)
+
+
+@runtime_checkable
+class TraceSink(Protocol):
+    """
+    Protocol for trace event sinks.
+    
+    Implementations receive action events and handle them
+    (e.g., print to stdout, write to file, send to server).
+    """
+    
+    def emit(self, event: ActionEvent) -> None:
+        """Emit an action event."""
+        ...
+    
+    def flush(self) -> None:
+        """Flush any buffered events."""
+        ...
+    
+    def close(self) -> None:
+        """Close the sink and release resources."""
+        ...
+
+
+class NoOpSink:
+    """
+    No-operation sink that discards all events.
+    
+    Used as the default sink when tracing is disabled.
+    Has near-zero overhead.
+    """
+    
+    __slots__ = ()
+    
+    def emit(self, event: ActionEvent) -> None:
+        """Discard the event."""
+        pass
+    
+    def flush(self) -> None:
+        """No-op."""
+        pass
+    
+    def close(self) -> None:
+        """No-op."""
+        pass
+
+
+class ListSink:
+    """
+    Sink that collects events in a list.
+    
+    Useful for testing and programmatic access to trace events.
+    """
+    
+    __slots__ = ("_events",)
+    
+    def __init__(self):
+        self._events: List[ActionEvent] = []
+    
+    def emit(self, event: ActionEvent) -> None:
+        """Add event to the list."""
+        self._events.append(event)
+    
+    def flush(self) -> None:
+        """No-op for list sink."""
+        pass
+    
+    def close(self) -> None:
+        """No-op for list sink."""
+        pass
+    
+    def get_events(self) -> List[ActionEvent]:
+        """Get all collected events."""
+        return self._events.copy()
+    
+    def clear(self) -> None:
+        """Clear all collected events."""
+        self._events.clear()
+
+
+@dataclass
+class ActionTraceConfig:
+    """
+    Configuration for action tracing.
+    
+    Usage:
+        # Default (disabled)
+        config = ActionTraceConfig()
+        
+        # Enable with stdout
+        config = ActionTraceConfig(enabled=True, sink_type="stdout")
+        
+        # Enable with JSONL file
+        config = ActionTraceConfig(
+            enabled=True,
+            sink_type="jsonl",
+            file_path="trace.jsonl",
+        )
+    """
+    enabled: bool = True
+    redact: bool = True
+    compact: bool = False
+    sink_type: str = "noop"  # noop, stdout, jsonl, file
+    file_path: Optional[str] = None
+    
+    def to_dict(self) -> Dict[str, Any]:
+        """Convert to dictionary."""
+        return {
+            "enabled": self.enabled,
+            "redact": self.redact,
+            "compact": self.compact,
+            "sink_type": self.sink_type,
+            "file_path": self.file_path,
+        }
+
+
+class TraceEmitter:
+    """
+    Emitter for action trace events.
+    
+    Provides a high-level API for emitting trace events.
+    Handles redaction and sink management.
+    
+    Usage:
+        sink = ListSink()
+        emitter = TraceEmitter(sink=sink)
+        
+        emitter.agent_start("researcher")
+        emitter.tool_start("search", {"query": "AI"})
+        emitter.tool_end("search", duration_ms=100, status="ok")
+        emitter.agent_end("researcher", duration_ms=500)
+    """
+    
+    __slots__ = ("_sink", "_enabled", "_redact", "_agent_starts")
+    
+    def __init__(
+        self,
+        sink: Optional[TraceSink] = None,
+        enabled: bool = True,
+        redact: bool = True,
+    ):
+        self._sink = sink or NoOpSink()
+        self._enabled = enabled
+        self._redact = redact
+        self._agent_starts: Dict[str, float] = {}  # Track agent start times
+    
+    def _emit(self, event: ActionEvent) -> None:
+        """Internal emit with enabled check."""
+        if not self._enabled:
+            return
+        self._sink.emit(event)
+    
+    def _redact_args(self, args: Optional[Dict[str, Any]]) -> Optional[Dict[str, Any]]:
+        """Redact sensitive data from args if enabled."""
+        if args is None or not self._redact:
+            return args
+        
+        from .redact import redact_dict
+        return redact_dict(args)
+    
+    def agent_start(
+        self,
+        agent_name: str,
+        agent_id: Optional[str] = None,
+        metadata: Optional[Dict[str, Any]] = None,
+    ) -> None:
+        """Emit agent start event."""
+        ts = time.time()
+        self._agent_starts[agent_name] = ts
+        
+        self._emit(ActionEvent(
+            event_type=ActionEventType.AGENT_START.value,
+            timestamp=ts,
+            agent_name=agent_name,
+            agent_id=agent_id,
+            metadata=metadata or {},
+        ))
+    
+    def agent_end(
+        self,
+        agent_name: str,
+        agent_id: Optional[str] = None,
+        duration_ms: Optional[float] = None,
+        status: str = "ok",
+        metadata: Optional[Dict[str, Any]] = None,
+    ) -> None:
+        """Emit agent end event."""
+        ts = time.time()
+        
+        # Calculate duration if not provided
+        if duration_ms is None and agent_name in self._agent_starts:
+            start_ts = self._agent_starts.pop(agent_name, None)
+            if start_ts:
+                duration_ms = (ts - start_ts) * 1000
+        
+        self._emit(ActionEvent(
+            event_type=ActionEventType.AGENT_END.value,
+            timestamp=ts,
+            agent_name=agent_name,
+            agent_id=agent_id,
+            duration_ms=duration_ms,
+            status=status,
+            metadata=metadata or {},
+        ))
+    
+    def tool_start(
+        self,
+        tool_name: str,
+        tool_args: Optional[Dict[str, Any]] = None,
+        agent_name: Optional[str] = None,
+        agent_id: Optional[str] = None,
+        metadata: Optional[Dict[str, Any]] = None,
+    ) -> None:
+        """Emit tool start event."""
+        self._emit(ActionEvent(
+            event_type=ActionEventType.TOOL_START.value,
+            timestamp=time.time(),
+            tool_name=tool_name,
+            tool_args=self._redact_args(tool_args),
+            agent_name=agent_name,
+            agent_id=agent_id,
+            metadata=metadata or {},
+        ))
+    
+    def tool_end(
+        self,
+        tool_name: str,
+        duration_ms: float,
+        status: str = "ok",
+        result_summary: Optional[str] = None,
+        agent_name: Optional[str] = None,
+        agent_id: Optional[str] = None,
+        error_message: Optional[str] = None,
+        metadata: Optional[Dict[str, Any]] = None,
+    ) -> None:
+        """Emit tool end event."""
+        self._emit(ActionEvent(
+            event_type=ActionEventType.TOOL_END.value,
+            timestamp=time.time(),
+            tool_name=tool_name,
+            duration_ms=duration_ms,
+            status=status,
+            tool_result_summary=result_summary,
+            agent_name=agent_name,
+            agent_id=agent_id,
+            error_message=error_message,
+            metadata=metadata or {},
+        ))
+    
+    def error(
+        self,
+        message: str,
+        tool_name: Optional[str] = None,
+        agent_name: Optional[str] = None,
+        agent_id: Optional[str] = None,
+        metadata: Optional[Dict[str, Any]] = None,
+    ) -> None:
+        """Emit error event."""
+        self._emit(ActionEvent(
+            event_type=ActionEventType.ERROR.value,
+            timestamp=time.time(),
+            error_message=message,
+            tool_name=tool_name,
+            agent_name=agent_name,
+            agent_id=agent_id,
+            status="error",
+            metadata=metadata or {},
+        ))
+    
+    def output(
+        self,
+        content: str,
+        agent_name: Optional[str] = None,
+        agent_id: Optional[str] = None,
+        metadata: Optional[Dict[str, Any]] = None,
+    ) -> None:
+        """Emit final output event."""
+        # Truncate long output for summary
+        summary = content[:500] + "..." if len(content) > 500 else content
+        
+        self._emit(ActionEvent(
+            event_type=ActionEventType.OUTPUT.value,
+            timestamp=time.time(),
+            tool_result_summary=summary,
+            agent_name=agent_name,
+            agent_id=agent_id,
+            metadata=metadata or {},
+        ))
+    
+    def flush(self) -> None:
+        """Flush the sink."""
+        self._sink.flush()
+    
+    def close(self) -> None:
+        """Close the sink."""
+        self._sink.close()
+
+
+# Global default emitter (NoOp by default)
+_default_emitter: Optional[TraceEmitter] = None
+
+
+def get_default_emitter() -> TraceEmitter:
+    """Get the default trace emitter."""
+    global _default_emitter
+    if _default_emitter is None:
+        _default_emitter = TraceEmitter(sink=NoOpSink(), enabled=False)
+    return _default_emitter
+
+
+def set_default_emitter(emitter: TraceEmitter) -> None:
+    """Set the default trace emitter."""
+    global _default_emitter
+    _default_emitter = emitter

src/praisonai-agents/praisonaiagents/trace/redact.py

@@ -0,0 +1,149 @@
+"""
+Redaction utilities for action trace.
+
+Provides deterministic redaction of sensitive data in tool arguments
+and results. Uses only stdlib for zero dependencies.
+"""
+
+from typing import Any, Dict, Set
+import re
+
+# Keys that should always be redacted (case-insensitive matching)
+REDACT_KEYS: Set[str] = {
+    # API keys and tokens
+    "api_key",
+    "apikey",
+    "api-key",
+    "secret",
+    "secret_key",
+    "secretkey",
+    "private_key",
+    "privatekey",
+    # Authentication
+    "password",
+    "passwd",
+    "pwd",
+    "token",
+    "bearer",
+    "authorization",
+    "auth",
+    "auth_token",
+    "access_token",
+    "refresh_token",
+    # Session and cookies
+    "cookie",
+    "session",
+    "session_id",
+    "sessionid",
+    "credential",
+    "credentials",
+    # Database
+    "connection_string",
+    "connectionstring",
+    "database_url",
+    "db_url",
+    "dburl",
+    "db_password",
+    "db_pass",
+    # Cloud provider specific
+    "aws_secret",
+    "aws_secret_access_key",
+    "azure_key",
+    "gcp_key",
+    "openai_api_key",
+    "anthropic_api_key",
+}
+
+# Compiled regex for case-insensitive matching
+_REDACT_PATTERN = re.compile(
+    r"(" + "|".join(re.escape(k) for k in REDACT_KEYS) + r")",
+    re.IGNORECASE
+)
+
+REDACTED_VALUE = "[REDACTED]"
+
+
+def _should_redact(key: str) -> bool:
+    """Check if a key should be redacted."""
+    # Normalize key for comparison
+    normalized = key.lower().replace("-", "_").replace(" ", "_")
+    
+    # Check exact match first
+    if normalized in REDACT_KEYS:
+        return True
+    
+    # Check if any redact key is contained in the normalized key
+    for redact_key in REDACT_KEYS:
+        if redact_key in normalized:
+            return True
+    
+    return False
+
+
+def redact_dict(data: Dict[str, Any], enabled: bool = True) -> Dict[str, Any]:
+    """
+    Redact sensitive values from a dictionary.
+    
+    Args:
+        data: Dictionary to redact
+        enabled: If False, returns data unchanged
+        
+    Returns:
+        New dictionary with sensitive values redacted
+    """
+    if not enabled:
+        return data
+    
+    if not data:
+        return {}
+    
+    return _redact_value(data)
+
+
+def _redact_value(value: Any) -> Any:
+    """Recursively redact sensitive values."""
+    if isinstance(value, dict):
+        return {k: _redact_key_value(k, v) for k, v in value.items()}
+    elif isinstance(value, list):
+        return [_redact_value(item) for item in value]
+    else:
+        return value
+
+
+def _redact_key_value(key: str, value: Any) -> Any:
+    """Redact a value if its key is sensitive."""
+    if _should_redact(key):
+        return REDACTED_VALUE
+    
+    # Recursively process nested structures
+    return _redact_value(value)
+
+
+def redact_string(text: str, enabled: bool = True) -> str:
+    """
+    Redact potential secrets from a string.
+    
+    This is a best-effort function that looks for common patterns
+    like "api_key=xxx" or "password: xxx" in strings.
+    
+    Args:
+        text: String to redact
+        enabled: If False, returns text unchanged
+        
+    Returns:
+        String with potential secrets redacted
+    """
+    if not enabled or not text:
+        return text
+    
+    # Pattern for key=value or key: value
+    patterns = [
+        (r'(["\']?)(' + '|'.join(re.escape(k) for k in REDACT_KEYS) + r')(["\']?\s*[:=]\s*["\']?)([^"\'\s,}\]]+)', 
+         r'\1\2\3[REDACTED]'),
+    ]
+    
+    result = text
+    for pattern, replacement in patterns:
+        result = re.sub(pattern, replacement, result, flags=re.IGNORECASE)
+    
+    return result

src/praisonai-agents/profile-streaming-delay.py

@@ -0,0 +1,232 @@
+#!/usr/bin/env python3
+"""
+Detailed profiler to identify the exact source of streaming delay.
+Traces every step from agent.start() to HTTP request.
+"""
+
+import os
+import sys
+import time
+import functools
+
+os.environ['LOGLEVEL'] = 'DEBUG'
+
+# Microsecond precision timestamps
+def ts():
+    return time.perf_counter()
+
+start_time = ts()
+
+def log(msg):
+    elapsed = (ts() - start_time) * 1000
+    print(f"[{elapsed:8.2f}ms] {msg}")
+
+log("Script starting...")
+
+# ============================================================
+# PATCH 1: OpenAI client initialization
+# ============================================================
+import openai
+
+original_openai_init = openai.OpenAI.__init__
+
+@functools.wraps(original_openai_init)
+def traced_openai_init(self, *args, **kwargs):
+    t0 = ts()
+    result = original_openai_init(self, *args, **kwargs)
+    log(f"OpenAI.__init__() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+openai.OpenAI.__init__ = traced_openai_init
+
+# ============================================================
+# PATCH 2: OpenAI completions.create
+# ============================================================
+original_create = openai.resources.chat.completions.Completions.create
+
+@functools.wraps(original_create)
+def traced_create(self, *args, **kwargs):
+    log(f">>> OpenAI.create() CALLED - stream={kwargs.get('stream', False)}")
+    t0 = ts()
+    result = original_create(self, *args, **kwargs)
+    log(f"<<< OpenAI.create() RETURNED in {(ts()-t0)*1000:.2f}ms")
+    return result
+
+openai.resources.chat.completions.Completions.create = traced_create
+
+# ============================================================
+# PATCH 3: httpx.Client.send (actual HTTP request)
+# ============================================================
+import httpx
+
+original_send = httpx.Client.send
+
+@functools.wraps(original_send)
+def traced_send(self, request, *args, **kwargs):
+    log(f">>> HTTPX.send() {request.method} {request.url}")
+    t0 = ts()
+    result = original_send(self, request, *args, **kwargs)
+    log(f"<<< HTTPX.send() returned in {(ts()-t0)*1000:.2f}ms - status={result.status_code}")
+    return result
+
+httpx.Client.send = traced_send
+
+# ============================================================
+# PATCH 4: Rich Live display
+# ============================================================
+from rich.live import Live
+
+original_live_init = Live.__init__
+original_live_enter = Live.__enter__
+original_live_exit = Live.__exit__
+
+@functools.wraps(original_live_init)
+def traced_live_init(self, *args, **kwargs):
+    log(f">>> Live.__init__() called")
+    t0 = ts()
+    result = original_live_init(self, *args, **kwargs)
+    log(f"<<< Live.__init__() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+@functools.wraps(original_live_enter)
+def traced_live_enter(self):
+    log(f">>> Live.__enter__() called")
+    t0 = ts()
+    result = original_live_enter(self)
+    log(f"<<< Live.__enter__() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+@functools.wraps(original_live_exit)
+def traced_live_exit(self, *args):
+    log(f">>> Live.__exit__() called")
+    t0 = ts()
+    result = original_live_exit(self, *args)
+    log(f"<<< Live.__exit__() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+Live.__init__ = traced_live_init
+Live.__enter__ = traced_live_enter
+Live.__exit__ = traced_live_exit
+
+log("Patches applied, importing praisonaiagents...")
+
+# ============================================================
+# Import and trace Agent
+# ============================================================
+t0 = ts()
+from praisonaiagents import Agent
+log(f"Agent import took {(ts()-t0)*1000:.2f}ms")
+
+# ============================================================
+# PATCH 5: Agent methods
+# ============================================================
+from praisonaiagents.agent.agent import Agent as AgentClass
+
+original_chat_completion = AgentClass._chat_completion
+original_format_tools = AgentClass._format_tools_for_completion
+
+@functools.wraps(original_chat_completion)
+def traced_chat_completion(self, messages, temperature=1.0, tools=None, stream=True, 
+                           reasoning_steps=False, task_name=None, task_description=None, task_id=None):
+    log(f">>> Agent._chat_completion() stream={stream}, _using_custom_llm={self._using_custom_llm}")
+    t0 = ts()
+    result = original_chat_completion(self, messages, temperature, tools, stream, 
+                                      reasoning_steps, task_name, task_description, task_id)
+    log(f"<<< Agent._chat_completion() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+@functools.wraps(original_format_tools)
+def traced_format_tools(self, tools):
+    log(f">>> Agent._format_tools_for_completion() tools={tools}")
+    t0 = ts()
+    result = original_format_tools(self, tools)
+    log(f"<<< Agent._format_tools_for_completion() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+AgentClass._chat_completion = traced_chat_completion
+AgentClass._format_tools_for_completion = traced_format_tools
+
+# ============================================================
+# PATCH 6: OpenAIClient methods
+# ============================================================
+from praisonaiagents.llm.openai_client import OpenAIClient
+
+original_chat_with_tools = OpenAIClient.chat_completion_with_tools
+original_process_stream = OpenAIClient.process_stream_response
+original_create_completion = OpenAIClient.create_completion
+original_format_tools_client = OpenAIClient.format_tools
+
+@functools.wraps(original_chat_with_tools)
+def traced_chat_with_tools(self, *args, **kwargs):
+    log(f">>> OpenAIClient.chat_completion_with_tools() stream={kwargs.get('stream', True)}")
+    t0 = ts()
+    result = original_chat_with_tools(self, *args, **kwargs)
+    log(f"<<< OpenAIClient.chat_completion_with_tools() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+@functools.wraps(original_process_stream)
+def traced_process_stream(self, *args, **kwargs):
+    log(f">>> OpenAIClient.process_stream_response()")
+    t0 = ts()
+    result = original_process_stream(self, *args, **kwargs)
+    log(f"<<< OpenAIClient.process_stream_response() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+@functools.wraps(original_create_completion)
+def traced_create_completion(self, *args, **kwargs):
+    log(f">>> OpenAIClient.create_completion() stream={kwargs.get('stream', False)}")
+    t0 = ts()
+    result = original_create_completion(self, *args, **kwargs)
+    log(f"<<< OpenAIClient.create_completion() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+@functools.wraps(original_format_tools_client)
+def traced_format_tools_client(self, tools):
+    log(f">>> OpenAIClient.format_tools() tools count={len(tools) if tools else 0}")
+    t0 = ts()
+    result = original_format_tools_client(self, tools)
+    log(f"<<< OpenAIClient.format_tools() took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+OpenAIClient.chat_completion_with_tools = traced_chat_with_tools
+OpenAIClient.process_stream_response = traced_process_stream
+OpenAIClient.create_completion = traced_create_completion
+OpenAIClient.format_tools = traced_format_tools_client
+
+# ============================================================
+# PATCH 7: sync_client property (lazy init)
+# ============================================================
+original_sync_client_getter = OpenAIClient.sync_client.fget
+
+def traced_sync_client_getter(self):
+    log(f">>> OpenAIClient.sync_client property accessed")
+    t0 = ts()
+    result = original_sync_client_getter(self)
+    log(f"<<< OpenAIClient.sync_client took {(ts()-t0)*1000:.2f}ms")
+    return result
+
+OpenAIClient.sync_client = property(traced_sync_client_getter)
+
+log("All patches applied. Creating agent...")
+
+# ============================================================
+# RUN THE TEST
+# ============================================================
+t0 = ts()
+agent = Agent(
+    instructions="You are a helpful assistant"
+)
+log(f"Agent created in {(ts()-t0)*1000:.2f}ms")
+log(f"Agent.stream={agent.stream}, Agent.verbose={agent.verbose}")
+
+log("=" * 60)
+log("Starting agent.start()...")
+log("=" * 60)
+
+t0 = ts()
+result = agent.start("Say hello in one word")
+log(f"agent.start() completed in {(ts()-t0)*1000:.2f}ms")
+
+log("=" * 60)
+log("DONE")
+log("=" * 60)

src/praisonai-agents/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "praisonaiagents"
-version = "0.11.2"
+version = "0.11.6"
 description = "Praison AI agents for completing complex tasks with Self Reflection Agents"
 readme = "README.md"
 requires-python = ">=3.10"

src/praisonai-agents/tests/benchmarks/output_mode_benchmark.py

@@ -0,0 +1,207 @@
+#!/usr/bin/env python3
+"""
+Micro-benchmark for output mode overhead comparison.
+
+Measures:
+1. Import-time cost (cold import)
+2. Agent initialization overhead per output mode
+3. Hot-path overhead (simulated callback execution)
+
+Usage:
+    python tests/benchmarks/output_mode_benchmark.py
+
+All stdlib - no external dependencies required.
+"""
+
+import subprocess
+import sys
+import time
+from statistics import mean, stdev
+
+
+def measure_import_time(mode: str, iterations: int = 5) -> dict:
+    """Measure cold import time for a specific output mode."""
+    times = []
+    
+    # Script that imports and creates an agent with specific output mode
+    script = f'''
+import time
+start = time.perf_counter()
+import os
+os.environ['PRAISONAI_OUTPUT'] = '{mode}'
+from praisonaiagents import Agent
+agent = Agent(name="BenchAgent", instructions="Test", output="{mode}")
+end = time.perf_counter()
+print(f"{{(end - start) * 1000:.3f}}")
+'''
+    
+    for _ in range(iterations):
+        result = subprocess.run(
+            [sys.executable, '-c', script],
+            capture_output=True,
+            text=True,
+            env={**dict(__import__('os').environ), 'PYTHONDONTWRITEBYTECODE': '1'}
+        )
+        if result.returncode == 0:
+            try:
+                times.append(float(result.stdout.strip()))
+            except ValueError:
+                pass
+    
+    if not times:
+        return {"mode": mode, "mean_ms": -1, "stdev_ms": -1, "samples": 0}
+    
+    return {
+        "mode": mode,
+        "mean_ms": mean(times),
+        "stdev_ms": stdev(times) if len(times) > 1 else 0,
+        "samples": len(times),
+    }
+
+
+def measure_callback_overhead(iterations: int = 1000) -> dict:
+    """Measure hot-path overhead of callback execution vs no callbacks."""
+    results = {}
+    
+    # Silent mode - no callbacks
+    silent_times = []
+    for _ in range(iterations):
+        start = time.perf_counter()
+        # Simulate what happens in silent mode: nothing
+        pass
+        end = time.perf_counter()
+        silent_times.append((end - start) * 1_000_000)  # microseconds
+    
+    results["silent"] = {
+        "mean_us": mean(silent_times),
+        "stdev_us": stdev(silent_times) if len(silent_times) > 1 else 0,
+    }
+    
+    # Actions mode - simulate callback overhead
+    actions_times = []
+    
+    # Simulate the work done by actions mode per tool call
+    def simulate_actions_callback():
+        import time as t
+        # Simulate: time.time(), dict creation, string formatting, stderr write
+        _ = t.time()
+        event = {
+            "type": "tool_call",
+            "tool": "test_tool",
+            "args": {"arg1": "value1", "arg2": "value2"},
+            "timestamp": t.time(),
+        }
+        formatted = f"[tool_call] {event['tool']}({event['args']})"
+        # Don't actually write to avoid polluting output
+        _ = formatted.encode()
+    
+    for _ in range(iterations):
+        start = time.perf_counter()
+        simulate_actions_callback()
+        end = time.perf_counter()
+        actions_times.append((end - start) * 1_000_000)  # microseconds
+    
+    results["actions"] = {
+        "mean_us": mean(actions_times),
+        "stdev_us": stdev(actions_times) if len(actions_times) > 1 else 0,
+    }
+    
+    return results
+
+
+def measure_agent_init_overhead() -> dict:
+    """Measure Agent.__init__ overhead for different output modes."""
+    import os
+    
+    # Clear any env var
+    os.environ.pop('PRAISONAI_OUTPUT', None)
+    
+    results = {}
+    iterations = 10
+    
+    for mode in ["silent", "actions", "verbose"]:
+        times = []
+        for _ in range(iterations):
+            # Import fresh each time to avoid caching effects
+            from praisonaiagents import Agent
+            
+            start = time.perf_counter()
+            agent = Agent(
+                name="BenchAgent",
+                instructions="Test agent",
+                output=mode,
+            )
+            end = time.perf_counter()
+            times.append((end - start) * 1000)  # milliseconds
+            del agent
+        
+        results[mode] = {
+            "mean_ms": mean(times),
+            "stdev_ms": stdev(times) if len(times) > 1 else 0,
+        }
+    
+    return results
+
+
+def main():
+    print("=" * 60)
+    print("OUTPUT MODE MICRO-BENCHMARK")
+    print("=" * 60)
+    print()
+    
+    # 1. Import time comparison
+    print("1. IMPORT + INIT TIME (cold start, subprocess)")
+    print("-" * 40)
+    
+    modes = ["silent", "actions", "verbose"]
+    import_results = []
+    
+    for mode in modes:
+        print(f"   Measuring {mode}...", end=" ", flush=True)
+        result = measure_import_time(mode, iterations=3)
+        import_results.append(result)
+        print(f"{result['mean_ms']:.1f}ms (±{result['stdev_ms']:.1f})")
+    
+    print()
+    
+    # 2. Agent init overhead (warm)
+    print("2. AGENT INIT OVERHEAD (warm, same process)")
+    print("-" * 40)
+    
+    init_results = measure_agent_init_overhead()
+    for mode, data in init_results.items():
+        print(f"   {mode:12s}: {data['mean_ms']:.2f}ms (±{data['stdev_ms']:.2f})")
+    
+    print()
+    
+    # 3. Hot-path callback overhead
+    print("3. HOT-PATH CALLBACK OVERHEAD (per call)")
+    print("-" * 40)
+    
+    callback_results = measure_callback_overhead(iterations=10000)
+    for mode, data in callback_results.items():
+        print(f"   {mode:12s}: {data['mean_us']:.3f}µs (±{data['stdev_us']:.3f})")
+    
+    print()
+    
+    # Summary
+    print("=" * 60)
+    print("SUMMARY")
+    print("=" * 60)
+    
+    silent_init = init_results.get("silent", {}).get("mean_ms", 0)
+    actions_init = init_results.get("actions", {}).get("mean_ms", 0)
+    
+    silent_callback = callback_results.get("silent", {}).get("mean_us", 0)
+    actions_callback = callback_results.get("actions", {}).get("mean_us", 0)
+    
+    print(f"   Init overhead (actions vs silent): +{actions_init - silent_init:.2f}ms")
+    print(f"   Per-call overhead (actions vs silent): +{actions_callback - silent_callback:.3f}µs")
+    print()
+    print("   DEFAULT MODE: silent (zero overhead)")
+    print("   OPT-IN: output='actions' for observability")
+    print()
+
+
+if __name__ == "__main__":
+    main()

src/praisonai-agents/tests/unit/agent/test_execution_api.py

@@ -0,0 +1,257 @@
+"""
+Tests for Agent execution API semantics: start(), run(), stream()
+
+TDD tests for the new execution API:
+- start() = interactive, streams by default in TTY
+- run() = production, silent by default
+- stream() = app-friendly iterator
+"""
+
+import pytest
+import sys
+from unittest.mock import patch
+
+
+class TestAgentStartMethod:
+    """Tests for Agent.start() - interactive, streaming by default in TTY"""
+    
+    def test_start_streams_by_default_in_tty(self):
+        """start() should stream by default when stdout is a TTY"""
+        from praisonaiagents import Agent
+        
+        # Create agent without stream attribute set (None)
+        agent = Agent(name="Test", instructions="You are helpful")
+        # Ensure stream attribute is None so TTY detection kicks in
+        agent.stream = None
+        
+        # Mock TTY detection
+        with patch.object(sys.stdout, 'isatty', return_value=True):
+            # Mock chat to avoid actual LLM call
+            with patch.object(agent, '_start_stream') as mock_stream:
+                mock_stream.return_value = iter(["Hello", " World"])
+                with patch.object(agent, 'chat') as mock_chat:
+                    agent.start("Hello")
+                    # Should call _start_stream when TTY
+                    mock_stream.assert_called_once()
+                    mock_chat.assert_not_called()
+    
+    def test_start_does_not_stream_when_not_tty(self):
+        """start() should not stream when stdout is not a TTY"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        # Mock non-TTY
+        with patch.object(sys.stdout, 'isatty', return_value=False):
+            with patch.object(agent, '_start_stream') as mock_stream:
+                with patch.object(agent, 'chat', return_value="Hello World") as mock_chat:
+                    result = agent.start("Hello")
+                    # Should call chat, not _start_stream
+                    mock_chat.assert_called_once()
+                    mock_stream.assert_not_called()
+    
+    def test_start_respects_explicit_stream_true(self):
+        """start(stream=True) should always stream"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        # Even when not TTY, explicit stream=True should stream
+        with patch.object(sys.stdout, 'isatty', return_value=False):
+            with patch.object(agent, '_start_stream') as mock_stream:
+                mock_stream.return_value = iter(["Hello"])
+                result = agent.start("Hello", stream=True)
+                mock_stream.assert_called_once()
+    
+    def test_start_respects_explicit_stream_false(self):
+        """start(stream=False) should never stream"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        # Even when TTY, explicit stream=False should not stream
+        with patch.object(sys.stdout, 'isatty', return_value=True):
+            with patch.object(agent, 'chat', return_value="Hello") as mock_chat:
+                with patch.object(agent, '_start_stream') as mock_stream:
+                    result = agent.start("Hello", stream=False)
+                    mock_chat.assert_called_once()
+                    mock_stream.assert_not_called()
+    
+    def test_start_respects_agent_stream_attribute(self):
+        """start() should respect agent's stream attribute over TTY detection"""
+        from praisonaiagents import Agent
+        
+        # Agent with stream=True should always stream
+        agent = Agent(name="Test", instructions="You are helpful", output="stream")
+        
+        with patch.object(sys.stdout, 'isatty', return_value=False):
+            with patch.object(agent, '_start_stream') as mock_stream:
+                mock_stream.return_value = iter(["Hello"])
+                # Agent's stream attribute should take precedence
+                if agent.stream:
+                    result = agent.start("Hello")
+                    mock_stream.assert_called()
+
+
+class TestAgentRunMethod:
+    """Tests for Agent.run() - production, silent by default"""
+    
+    def test_run_does_not_stream_by_default(self):
+        """run() should not stream by default"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        with patch.object(agent, 'chat', return_value="Hello World") as mock_chat:
+            with patch.object(agent, '_start_stream') as mock_stream:
+                result = agent.run("Hello")
+                # Should call chat with stream=False
+                mock_chat.assert_called_once()
+                call_kwargs = mock_chat.call_args[1]
+                assert call_kwargs.get('stream') == False
+                mock_stream.assert_not_called()
+    
+    def test_run_returns_result_directly(self):
+        """run() should return the result directly, not a generator"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        with patch.object(agent, 'chat', return_value="Hello World") as mock_chat:
+            result = agent.run("Hello")
+            assert result == "Hello World"
+            assert not hasattr(result, '__iter__') or isinstance(result, str)
+    
+    def test_run_ignores_tty_detection(self):
+        """run() should ignore TTY detection and never stream by default"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        # Even in TTY, run() should not stream
+        with patch.object(sys.stdout, 'isatty', return_value=True):
+            with patch.object(agent, 'chat', return_value="Hello") as mock_chat:
+                with patch.object(agent, '_start_stream') as mock_stream:
+                    result = agent.run("Hello")
+                    mock_chat.assert_called_once()
+                    mock_stream.assert_not_called()
+    
+    def test_run_can_be_forced_to_stream(self):
+        """run(stream=True) should stream when explicitly requested"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        with patch.object(agent, 'chat', return_value="Hello") as mock_chat:
+            result = agent.run("Hello", stream=True)
+            # When stream=True is passed, it should be passed to chat
+            call_kwargs = mock_chat.call_args[1]
+            assert call_kwargs.get('stream') == True
+
+
+class TestAgentIterStreamMethod:
+    """Tests for Agent.iter_stream() - app-friendly iterator"""
+    
+    def test_iter_stream_returns_iterator(self):
+        """iter_stream() should return an iterator"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        with patch.object(agent, '_start_stream') as mock_stream:
+            mock_stream.return_value = iter(["Hello", " ", "World"])
+            result = agent.iter_stream("Hello")
+            # Should be an iterator/generator
+            assert hasattr(result, '__iter__')
+            assert hasattr(result, '__next__')
+    
+    def test_iter_stream_yields_chunks(self):
+        """iter_stream() should yield response chunks"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        with patch.object(agent, '_start_stream') as mock_stream:
+            mock_stream.return_value = iter(["Hello", " ", "World"])
+            chunks = list(agent.iter_stream("Hello"))
+            assert chunks == ["Hello", " ", "World"]
+    
+    def test_iter_stream_always_streams(self):
+        """iter_stream() should always stream regardless of TTY"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        # Even when not TTY
+        with patch.object(sys.stdout, 'isatty', return_value=False):
+            with patch.object(agent, '_start_stream') as mock_stream:
+                mock_stream.return_value = iter(["Hello"])
+                result = agent.iter_stream("Hello")
+                # Consume the generator
+                list(result)
+                mock_stream.assert_called_once()
+
+
+class TestAsyncMethods:
+    """Tests for async versions: arun(), astart()"""
+    
+    @pytest.mark.asyncio
+    async def test_arun_does_not_stream_by_default(self):
+        """arun() should not stream by default"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        
+        with patch.object(agent, 'achat', return_value="Hello World") as mock_achat:
+            result = await agent.arun("Hello")
+            mock_achat.assert_called_once()
+            call_kwargs = mock_achat.call_args[1]
+            assert call_kwargs.get('stream') == False
+    
+    @pytest.mark.asyncio
+    async def test_astart_streams_in_tty(self):
+        """astart() should stream when in TTY"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        # Ensure stream attribute is None so TTY detection kicks in
+        agent.stream = None
+        
+        with patch.object(sys.stdout, 'isatty', return_value=True):
+            with patch.object(agent, 'achat', return_value="Hello") as mock_achat:
+                await agent.astart("Hello")
+                call_kwargs = mock_achat.call_args[1]
+                assert call_kwargs.get('stream')
+
+
+class TestMethodSemanticDistinction:
+    """Tests to verify start() and run() have distinct semantics"""
+    
+    def test_start_and_run_are_different(self):
+        """start() and run() should have different default behaviors"""
+        from praisonaiagents import Agent
+        
+        agent = Agent(name="Test", instructions="You are helpful")
+        # Ensure stream attribute is None so TTY detection kicks in
+        agent.stream = None
+        
+        # In TTY context
+        with patch.object(sys.stdout, 'isatty', return_value=True):
+            with patch.object(agent, 'chat', return_value="Hello") as mock_chat:
+                with patch.object(agent, '_start_stream') as mock_stream:
+                    mock_stream.return_value = iter(["Hello"])
+                    
+                    # start() should stream
+                    agent.start("Hello")
+                    start_streamed = mock_stream.called
+                    
+                    mock_stream.reset_mock()
+                    mock_chat.reset_mock()
+                    
+                    # run() should not stream
+                    agent.run("Hello")
+                    run_streamed = mock_stream.called
+                    
+                    # They should behave differently
+                    assert start_streamed
+                    assert not run_streamed

src/praisonai-agents/tests/unit/trace/__init__.py

@@ -0,0 +1 @@
+"""Tests for the trace module."""

src/praisonai-agents/tests/unit/trace/test_protocol.py

@@ -0,0 +1,450 @@
+"""
+Tests for the trace protocol module.
+
+TDD tests for ActionEvent, TraceSink protocol, and redaction.
+"""
+
+import time
+import json
+
+
+class TestActionEvent:
+    """Tests for ActionEvent dataclass."""
+    
+    def test_action_event_creation_minimal(self):
+        """Test creating ActionEvent with minimal fields."""
+        from praisonaiagents.trace.protocol import ActionEvent
+        
+        event = ActionEvent(
+            event_type="agent_start",
+            timestamp=1736665496.123,
+        )
+        
+        assert event.event_type == "agent_start"
+        assert event.timestamp == 1736665496.123
+        assert event.schema_version == "1.0"
+        assert event.agent_id is None
+        assert event.agent_name is None
+        assert event.tool_name is None
+    
+    def test_action_event_creation_full(self):
+        """Test creating ActionEvent with all fields."""
+        from praisonaiagents.trace.protocol import ActionEvent
+        
+        event = ActionEvent(
+            event_type="tool_end",
+            timestamp=1736665497.456,
+            agent_id="agent-123",
+            agent_name="researcher",
+            tool_name="web_search",
+            tool_args={"query": "AI trends"},
+            tool_result_summary="Found 10 results",
+            duration_ms=800.5,
+            status="ok",
+            error_message=None,
+            metadata={"provider": "duckduckgo"},
+        )
+        
+        assert event.event_type == "tool_end"
+        assert event.agent_name == "researcher"
+        assert event.tool_name == "web_search"
+        assert event.tool_args == {"query": "AI trends"}
+        assert event.duration_ms == 800.5
+        assert event.status == "ok"
+        assert event.metadata == {"provider": "duckduckgo"}
+    
+    def test_action_event_to_dict(self):
+        """Test ActionEvent.to_dict() method."""
+        from praisonaiagents.trace.protocol import ActionEvent
+        
+        event = ActionEvent(
+            event_type="agent_start",
+            timestamp=1736665496.123,
+            agent_name="writer",
+        )
+        
+        d = event.to_dict()
+        
+        assert d["schema_version"] == "1.0"
+        assert d["event_type"] == "agent_start"
+        assert d["timestamp"] == 1736665496.123
+        assert d["agent_name"] == "writer"
+        # None values should be excluded
+        assert "tool_name" not in d or d["tool_name"] is None
+    
+    def test_action_event_to_json(self):
+        """Test ActionEvent.to_json() method."""
+        from praisonaiagents.trace.protocol import ActionEvent
+        
+        event = ActionEvent(
+            event_type="tool_start",
+            timestamp=1736665497.0,
+            tool_name="read_file",
+            tool_args={"path": "/tmp/test.txt"},
+        )
+        
+        json_str = event.to_json()
+        parsed = json.loads(json_str)
+        
+        assert parsed["event_type"] == "tool_start"
+        assert parsed["tool_name"] == "read_file"
+        assert parsed["tool_args"] == {"path": "/tmp/test.txt"}
+    
+    def test_action_event_types(self):
+        """Test all valid event types."""
+        from praisonaiagents.trace.protocol import ActionEvent, ActionEventType
+        
+        valid_types = [
+            ActionEventType.AGENT_START,
+            ActionEventType.AGENT_END,
+            ActionEventType.TOOL_START,
+            ActionEventType.TOOL_END,
+            ActionEventType.ERROR,
+            ActionEventType.OUTPUT,
+        ]
+        
+        for event_type in valid_types:
+            event = ActionEvent(
+                event_type=event_type.value,
+                timestamp=time.time(),
+            )
+            assert event.event_type == event_type.value
+
+
+class TestRedaction:
+    """Tests for redaction functionality."""
+    
+    def test_redact_api_key(self):
+        """Test redacting api_key field."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        data = {"api_key": "sk-1234567890abcdef", "query": "hello"}
+        redacted = redact_dict(data)
+        
+        assert redacted["api_key"] == "[REDACTED]"
+        assert redacted["query"] == "hello"
+    
+    def test_redact_multiple_sensitive_keys(self):
+        """Test redacting multiple sensitive keys."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        data = {
+            "api_key": "sk-secret",
+            "password": "mypassword123",
+            "token": "bearer-token",
+            "authorization": "Bearer xyz",
+            "secret": "top-secret",
+            "normal_field": "visible",
+        }
+        redacted = redact_dict(data)
+        
+        assert redacted["api_key"] == "[REDACTED]"
+        assert redacted["password"] == "[REDACTED]"
+        assert redacted["token"] == "[REDACTED]"
+        assert redacted["authorization"] == "[REDACTED]"
+        assert redacted["secret"] == "[REDACTED]"
+        assert redacted["normal_field"] == "visible"
+    
+    def test_redact_nested_dict(self):
+        """Test redacting nested dictionaries."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        data = {
+            "config": {
+                "api_key": "sk-nested",
+                "endpoint": "https://api.example.com",
+            },
+            "name": "test",
+        }
+        redacted = redact_dict(data)
+        
+        assert redacted["config"]["api_key"] == "[REDACTED]"
+        assert redacted["config"]["endpoint"] == "https://api.example.com"
+        assert redacted["name"] == "test"
+    
+    def test_redact_list_of_dicts(self):
+        """Test redacting list containing dictionaries."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        data = {
+            "items": [
+                {"name": "item1", "secret": "hidden1"},
+                {"name": "item2", "secret": "hidden2"},
+            ]
+        }
+        redacted = redact_dict(data)
+        
+        assert redacted["items"][0]["name"] == "item1"
+        assert redacted["items"][0]["secret"] == "[REDACTED]"
+        assert redacted["items"][1]["secret"] == "[REDACTED]"
+    
+    def test_redact_case_insensitive(self):
+        """Test that redaction is case-insensitive."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        data = {
+            "API_KEY": "key1",
+            "Api_Key": "key2",
+            "apiKey": "key3",
+            "APIKEY": "key4",
+        }
+        redacted = redact_dict(data)
+        
+        assert redacted["API_KEY"] == "[REDACTED]"
+        assert redacted["Api_Key"] == "[REDACTED]"
+        assert redacted["apiKey"] == "[REDACTED]"
+        assert redacted["APIKEY"] == "[REDACTED]"
+    
+    def test_redact_connection_strings(self):
+        """Test redacting database connection strings."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        data = {
+            "connection_string": "postgresql://user:pass@host/db",
+            "database_url": "mysql://root:secret@localhost/mydb",
+            "db_url": "redis://default:password@redis:6379",
+        }
+        redacted = redact_dict(data)
+        
+        assert redacted["connection_string"] == "[REDACTED]"
+        assert redacted["database_url"] == "[REDACTED]"
+        assert redacted["db_url"] == "[REDACTED]"
+    
+    def test_redact_preserves_original(self):
+        """Test that redaction doesn't modify original dict."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        original = {"api_key": "secret", "name": "test"}
+        redacted = redact_dict(original)
+        
+        assert original["api_key"] == "secret"
+        assert redacted["api_key"] == "[REDACTED]"
+    
+    def test_redact_empty_dict(self):
+        """Test redacting empty dictionary."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        assert redact_dict({}) == {}
+    
+    def test_redact_none_values(self):
+        """Test handling None values."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        data = {"api_key": None, "name": "test"}
+        redacted = redact_dict(data)
+        
+        # None values for sensitive keys should still be redacted
+        assert redacted["api_key"] == "[REDACTED]"
+        assert redacted["name"] == "test"
+    
+    def test_redact_disabled(self):
+        """Test disabling redaction."""
+        from praisonaiagents.trace.redact import redact_dict
+        
+        data = {"api_key": "visible", "name": "test"}
+        result = redact_dict(data, enabled=False)
+        
+        assert result["api_key"] == "visible"
+        assert result["name"] == "test"
+
+
+class TestTraceSink:
+    """Tests for TraceSink protocol."""
+    
+    def test_noop_sink(self):
+        """Test NoOpSink does nothing."""
+        from praisonaiagents.trace.protocol import NoOpSink, ActionEvent
+        
+        sink = NoOpSink()
+        event = ActionEvent(event_type="agent_start", timestamp=time.time())
+        
+        # Should not raise
+        sink.emit(event)
+        sink.flush()
+        sink.close()
+    
+    def test_list_sink(self):
+        """Test ListSink collects events."""
+        from praisonaiagents.trace.protocol import ListSink, ActionEvent
+        
+        sink = ListSink()
+        
+        event1 = ActionEvent(event_type="agent_start", timestamp=1.0)
+        event2 = ActionEvent(event_type="tool_start", timestamp=2.0)
+        
+        sink.emit(event1)
+        sink.emit(event2)
+        
+        events = sink.get_events()
+        assert len(events) == 2
+        assert events[0].event_type == "agent_start"
+        assert events[1].event_type == "tool_start"
+    
+    def test_list_sink_clear(self):
+        """Test ListSink.clear() method."""
+        from praisonaiagents.trace.protocol import ListSink, ActionEvent
+        
+        sink = ListSink()
+        sink.emit(ActionEvent(event_type="test", timestamp=1.0))
+        
+        assert len(sink.get_events()) == 1
+        
+        sink.clear()
+        
+        assert len(sink.get_events()) == 0
+
+
+class TestTraceEmitter:
+    """Tests for TraceEmitter class."""
+    
+    def test_emitter_with_noop_sink(self):
+        """Test TraceEmitter with NoOpSink."""
+        from praisonaiagents.trace.protocol import TraceEmitter, NoOpSink
+        
+        emitter = TraceEmitter(sink=NoOpSink())
+        
+        # Should not raise
+        emitter.agent_start("test-agent", agent_id="123")
+        emitter.tool_start("web_search", {"query": "test"})
+        emitter.tool_end("web_search", duration_ms=100, status="ok")
+        emitter.agent_end("test-agent", duration_ms=500)
+    
+    def test_emitter_with_list_sink(self):
+        """Test TraceEmitter collects events."""
+        from praisonaiagents.trace.protocol import TraceEmitter, ListSink
+        
+        sink = ListSink()
+        emitter = TraceEmitter(sink=sink)
+        
+        emitter.agent_start("researcher")
+        emitter.tool_start("search", {"q": "AI"})
+        emitter.tool_end("search", duration_ms=100, status="ok")
+        emitter.agent_end("researcher", duration_ms=200)
+        
+        events = sink.get_events()
+        assert len(events) == 4
+        assert events[0].event_type == "agent_start"
+        assert events[1].event_type == "tool_start"
+        assert events[2].event_type == "tool_end"
+        assert events[3].event_type == "agent_end"
+    
+    def test_emitter_redacts_by_default(self):
+        """Test TraceEmitter redacts sensitive data by default."""
+        from praisonaiagents.trace.protocol import TraceEmitter, ListSink
+        
+        sink = ListSink()
+        emitter = TraceEmitter(sink=sink, redact=True)
+        
+        emitter.tool_start("api_call", {"api_key": "sk-secret", "query": "test"})
+        
+        events = sink.get_events()
+        assert events[0].tool_args["api_key"] == "[REDACTED]"
+        assert events[0].tool_args["query"] == "test"
+    
+    def test_emitter_disabled(self):
+        """Test TraceEmitter when disabled."""
+        from praisonaiagents.trace.protocol import TraceEmitter, ListSink
+        
+        sink = ListSink()
+        emitter = TraceEmitter(sink=sink, enabled=False)
+        
+        emitter.agent_start("test")
+        emitter.tool_start("tool", {})
+        
+        # No events should be emitted
+        assert len(sink.get_events()) == 0
+    
+    def test_emitter_error_event(self):
+        """Test TraceEmitter error event."""
+        from praisonaiagents.trace.protocol import TraceEmitter, ListSink
+        
+        sink = ListSink()
+        emitter = TraceEmitter(sink=sink)
+        
+        emitter.error("Something went wrong", tool_name="failing_tool")
+        
+        events = sink.get_events()
+        assert len(events) == 1
+        assert events[0].event_type == "error"
+        assert events[0].error_message == "Something went wrong"
+        assert events[0].tool_name == "failing_tool"
+    
+    def test_emitter_output_event(self):
+        """Test TraceEmitter output event for final agent output."""
+        from praisonaiagents.trace.protocol import TraceEmitter, ListSink
+        
+        sink = ListSink()
+        emitter = TraceEmitter(sink=sink)
+        
+        emitter.output("The analysis shows three major trends...")
+        
+        events = sink.get_events()
+        assert len(events) == 1
+        assert events[0].event_type == "output"
+        assert "three major trends" in events[0].tool_result_summary
+
+
+class TestActionTraceConfig:
+    """Tests for ActionTraceConfig."""
+    
+    def test_config_defaults(self):
+        """Test ActionTraceConfig default values."""
+        from praisonaiagents.trace.protocol import ActionTraceConfig
+        
+        config = ActionTraceConfig()
+        
+        assert config.enabled is True
+        assert config.redact is True
+        assert config.compact is False
+        assert config.sink_type == "noop"
+        assert config.file_path is None
+    
+    def test_config_custom(self):
+        """Test ActionTraceConfig custom values."""
+        from praisonaiagents.trace.protocol import ActionTraceConfig
+        
+        config = ActionTraceConfig(
+            enabled=True,
+            redact=False,
+            compact=True,
+            sink_type="jsonl",
+            file_path="/tmp/trace.jsonl",
+        )
+        
+        assert config.redact is False
+        assert config.compact is True
+        assert config.sink_type == "jsonl"
+        assert config.file_path == "/tmp/trace.jsonl"
+
+
+class TestPerformance:
+    """Performance tests for trace module."""
+    
+    def test_noop_sink_overhead(self):
+        """Test that NoOpSink has minimal overhead."""
+        from praisonaiagents.trace.protocol import TraceEmitter, NoOpSink, ActionEvent
+        
+        emitter = TraceEmitter(sink=NoOpSink())
+        
+        start = time.perf_counter()
+        for _ in range(10000):
+            emitter.tool_start("test", {"arg": "value"})
+            emitter.tool_end("test", duration_ms=1, status="ok")
+        elapsed = time.perf_counter() - start
+        
+        # Should complete 10000 iterations in under 100ms
+        assert elapsed < 0.1, f"NoOpSink too slow: {elapsed:.3f}s for 10000 iterations"
+    
+    def test_disabled_emitter_overhead(self):
+        """Test that disabled emitter has near-zero overhead."""
+        from praisonaiagents.trace.protocol import TraceEmitter, ListSink
+        
+        emitter = TraceEmitter(sink=ListSink(), enabled=False)
+        
+        start = time.perf_counter()
+        for _ in range(10000):
+            emitter.tool_start("test", {"arg": "value"})
+        elapsed = time.perf_counter() - start
+        
+        # Should complete 10000 iterations in under 500ms (very conservative)
+        assert elapsed < 0.5, f"Disabled emitter too slow: {elapsed:.3f}s"

src/praisonai-agents/uv.lock

@@ -2422,7 +2422,7 @@ wheels = [
 
 [[package]]
 name = "praisonaiagents"
-version = "0.11.2"
+version = "0.11.6"
 source = { editable = "." }
 dependencies = [
     { name = "aiohttp" },

src/praisonai/praisonai.rb

@@ -3,8 +3,8 @@ class Praisonai < Formula
   
     desc "AI tools for various AI applications"
     homepage "https://github.com/MervinPraison/PraisonAI"
-    url "https://github.com/MervinPraison/PraisonAI/archive/refs/tags/v3.7.3.tar.gz"
-    sha256 `curl -sL https://github.com/MervinPraison/PraisonAI/archive/refs/tags/v3.7.3.tar.gz | shasum -a 256`.split.first
+    url "https://github.com/MervinPraison/PraisonAI/archive/refs/tags/v3.8.0.tar.gz"
+    sha256 `curl -sL https://github.com/MervinPraison/PraisonAI/archive/refs/tags/v3.8.0.tar.gz | shasum -a 256`.split.first
     license "MIT"
   
     depends_on "[email protected]"

src/praisonai/praisonai/cli/app.py

@@ -141,11 +141,18 @@ def main_callback(
     )
     set_output_controller(state.output_controller)
     
-    # If no command provided, show help or handle legacy behavior
+    # If no command provided, start interactive mode
     if ctx.invoked_subcommand is None:
-        # Check if there are remaining args (legacy behavior)
-        # This is handled by the legacy adapter in main.py
-        pass
+        from praisonai.cli.interactive.async_tui import AsyncTUI, AsyncTUIConfig
+        
+        tui_config = AsyncTUIConfig(
+            model="gpt-4o-mini",
+            show_logo=True,
+            show_status_bar=state.output_format != OutputFormat.json,
+        )
+        
+        tui = AsyncTUI(config=tui_config)
+        tui.run()
 
 
 def get_output_controller() -> OutputController:

src/praisonai/praisonai/cli/branding.py

@@ -0,0 +1,94 @@
+"""
+PraisonAI Branding - Single Source of Truth.
+
+This module provides unified branding assets (logo, version, product name)
+for all interactive UIs. All frontends should import from here.
+
+Usage:
+    from praisonai.cli.branding import get_logo, get_version, PRODUCT_NAME
+"""
+
+# Product name
+PRODUCT_NAME = "Praison AI"
+
+# ASCII Art Logos - responsive to terminal width
+LOGO_LARGE = r"""
+ ██████╗ ██████╗  █████╗ ██╗███████╗ ██████╗ ███╗   ██╗     █████╗ ██╗
+ ██╔══██╗██╔══██╗██╔══██╗██║██╔════╝██╔═══██╗████╗  ██║    ██╔══██╗██║
+ ██████╔╝██████╔╝███████║██║███████╗██║   ██║██╔██╗ ██║    ███████║██║
+ ██╔═══╝ ██╔══██╗██╔══██║██║╚════██║██║   ██║██║╚██╗██║    ██╔══██║██║
+ ██║     ██║  ██║██║  ██║██║███████║╚██████╔╝██║ ╚████║    ██║  ██║██║
+ ╚═╝     ╚═╝  ╚═╝╚═╝  ╚═╝╚═╝╚══════╝ ╚═════╝ ╚═╝  ╚═══╝    ╚═╝  ╚═╝╚═╝
+"""
+
+LOGO_MEDIUM = r"""
+ ╔═╗┬─┐┌─┐┬┌─┐┌─┐┌┐┌  ╔═╗╦
+ ╠═╝├┬┘├─┤│└─┐│ ││││  ╠═╣║
+ ╩  ┴└─┴ ┴┴└─┘└─┘┘└┘  ╩ ╩╩
+"""
+
+LOGO_SMALL = "▶ Praison AI"
+
+LOGO_MINIMAL = "Praison AI"
+
+
+def get_version() -> str:
+    """Get PraisonAI version string."""
+    try:
+        from praisonai import __version__
+        return __version__
+    except Exception:
+        return "1.0.0"
+
+
+def get_logo(width: int = 80) -> str:
+    """
+    Get appropriate logo based on terminal width.
+    
+    Args:
+        width: Terminal width in characters
+        
+    Returns:
+        ASCII art logo string appropriate for the width
+    """
+    if width >= 75:
+        return LOGO_LARGE
+    elif width >= 40:
+        return LOGO_MEDIUM
+    else:
+        return LOGO_SMALL
+
+
+def get_banner(width: int = 80, show_version: bool = True, model: str = None) -> str:
+    """
+    Get full banner with logo, version, and optional model info.
+    
+    Args:
+        width: Terminal width
+        show_version: Whether to include version
+        model: Optional model name to display
+        
+    Returns:
+        Complete banner string
+    """
+    lines = []
+    
+    # Logo
+    logo = get_logo(width)
+    lines.append(logo.strip())
+    lines.append("")
+    
+    # Version and model
+    version_line = f"  v{get_version()}"
+    if model:
+        version_line += f" · Model: {model}"
+    lines.append(version_line)
+    lines.append("")
+    
+    return "\n".join(lines)
+
+
+def get_welcome_tips() -> str:
+    """Get welcome tips for new sessions."""
+    return """  Type your message and press Enter. Use /help for commands.
+  Use PageUp/PageDown or Ctrl+Up/Down to scroll."""

src/praisonai/praisonai/cli/commands/chat.py

@@ -67,6 +67,7 @@ def chat_main(
     workspace: Optional[str] = typer.Option(None, "--workspace", "-w", help="Workspace directory"),
     no_acp: bool = typer.Option(False, "--no-acp", help="Disable ACP tools"),
     no_lsp: bool = typer.Option(False, "--no-lsp", help="Disable LSP tools"),
+    safe_mode: bool = typer.Option(False, "--safe", help="Safe mode: require approval for file writes and commands"),
     autonomy: bool = typer.Option(True, "--autonomy/--no-autonomy", help="Enable agent autonomy for complex tasks"),
     # NEW: Agent-like consolidated params for ALL GREEN consistency
     knowledge: Optional[str] = typer.Option(
@@ -107,8 +108,8 @@ def chat_main(
         flag_value="true",
     ),
     output: Optional[str] = typer.Option(
-        None, "--output",
-        help="Output preset. Use --output=verbose, --output=minimal, --output=silent",
+        None, "--output", "-o",
+        help="Output mode (default: actions). Options: actions, plain, verbose, json, silent",
     ),
     execution: Optional[str] = typer.Option(
         None, "--execution",
@@ -126,6 +127,7 @@ def chat_main(
     ),
     profile: bool = typer.Option(False, "--profile", help="Enable CLI profiling (timing breakdown)"),
     profile_deep: bool = typer.Option(False, "--profile-deep", help="Enable deep profiling (cProfile stats, higher overhead)"),
+    debug: bool = typer.Option(False, "--debug", help="Enable debug logging to ~/.praisonai/async_tui_debug.log"),
     # UI backend selection
     ui_backend: str = typer.Option("auto", "--ui-backend", help="UI backend: auto, plain, rich, mg (middle-ground)"),
     json_output: bool = typer.Option(False, "--json", help="Output JSON (forces plain backend)"),
@@ -149,25 +151,12 @@ def chat_main(
         praisonai chat "Summarize this" --file README.md
         praisonai chat "What is 2+2?" --profile
     """
-    import asyncio
     import os
     
     # Set workspace if provided
     if workspace:
         os.environ["PRAISONAI_WORKSPACE"] = workspace
     
-    # Configure UI backend
-    from praisonai.cli.ui import select_backend, UIConfig
-    ui_config = UIConfig(
-        ui_backend=ui_backend,
-        json_output=json_output,
-        no_color=no_color,
-        theme=theme,
-        compact=compact,
-        verbose=verbose,
-    )
-    backend = select_backend(ui_config)
-    
     # Handle profiling for single prompt mode
     if prompt and (profile or profile_deep):
         _run_profiled_chat(
@@ -184,60 +173,38 @@ def chat_main(
         typer.echo("   Use: praisonai chat \"your prompt\" --profile", err=True)
     
     # Parse memory flag: --no-memory takes precedence, then --memory value
-    memory_value = _parse_memory_flag(memory, no_memory)
+    # TODO: Pass memory_value to TUI when memory support is added
+    _parse_memory_flag(memory, no_memory)
     
-    # Try InteractiveCore first (preferred terminal-native implementation)
-    try:
-        from praisonai.cli.interactive import InteractiveCore, InteractiveConfig
-        from praisonai.cli.interactive.frontends import RichFrontend
-        
-        config = InteractiveConfig(
-            model=model,
-            session_id=session_id,
-            continue_session=continue_session,
-            workspace=workspace or None,
-            verbose=verbose,
-            memory=memory_value,
-            files=list(file) if file else [],
-            autonomy=autonomy,
-        )
-        
-        core = InteractiveCore(config=config)
-        
-        if prompt:
-            # Single prompt mode - use backend for output
-            from praisonai.cli.ui.events import UIEventType
-            
-            async def run_prompt():
-                if continue_session:
-                    core.continue_session()
-                
-                backend.emit(UIEventType.MESSAGE_START, {})
-                response = await core.prompt(prompt)
-                backend.emit(UIEventType.MESSAGE_CHUNK, {'content': response})
-                backend.emit(UIEventType.MESSAGE_END, {})
-            
-            asyncio.run(run_prompt())
-        else:
-            # Interactive REPL mode - pass ui_config to frontend
-            frontend = RichFrontend(core=core, config=config, ui_config=ui_config)
-            asyncio.run(frontend.run())
-            
-    except ImportError:
-        # Fallback to legacy terminal-native interactive mode
-        _run_legacy_terminal_chat(
-            prompt=prompt,
-            model=model,
-            verbose=verbose,
-            memory=memory,
-            tools=tools,
-            user_id=user_id,
-            session_id=session_id,
-            continue_session=continue_session,
-            workspace=workspace,
-            no_acp=no_acp,
-            no_lsp=no_lsp,
-        )
+    # Set approval mode based on --safe flag
+    import os
+    if safe_mode:
+        os.environ["PRAISON_APPROVAL_MODE"] = "prompt"
+    else:
+        os.environ["PRAISON_APPROVAL_MODE"] = "auto"
+    
+    # Use the async TUI (non-blocking, scrollable output)
+    from praisonai.cli.interactive.async_tui import AsyncTUI, AsyncTUIConfig
+    
+    tui_config = AsyncTUIConfig(
+        model=model or "gpt-4o-mini",
+        show_logo=not compact,
+        show_status_bar=not compact,
+        session_id=session_id,
+        workspace=workspace,
+        debug=debug,
+    )
+    
+    tui = AsyncTUI(config=tui_config)
+    
+    if prompt:
+        # Single prompt mode - direct response, no streaming
+        response = tui.run_single(prompt)
+        if response:
+            print(response)
+    else:
+        # Interactive split-pane TUI mode
+        tui.run()
 
 
 def _run_profiled_chat(

src/praisonai/praisonai/cli/commands/code.py

@@ -23,6 +23,7 @@ def code_main(
     file: Optional[List[str]] = typer.Option(None, "--file", "-f", help="Attach file(s) to context"),
     no_acp: bool = typer.Option(False, "--no-acp", help="Disable ACP tools (file operations)"),
     no_lsp: bool = typer.Option(False, "--no-lsp", help="Disable LSP tools (code intelligence)"),
+    safe_mode: bool = typer.Option(False, "--safe", help="Safe mode: require approval for file writes and commands"),
     session_id: Optional[str] = typer.Option(None, "--session", "-s", help="Session ID to resume"),
     continue_session: bool = typer.Option(False, "--continue", "-c", help="Continue last session"),
     autonomy: bool = typer.Option(True, "--autonomy/--no-autonomy", help="Enable agent autonomy for complex tasks"),
@@ -54,6 +55,12 @@ def code_main(
     # Enable code-specific environment
     os.environ["PRAISONAI_CODE_MODE"] = "true"
     
+    # Set approval mode based on --safe flag
+    if safe_mode:
+        os.environ["PRAISON_APPROVAL_MODE"] = "prompt"
+    else:
+        os.environ["PRAISON_APPROVAL_MODE"] = "auto"
+    
     # Handle profiling for single prompt mode
     if prompt and (profile or profile_deep):
         _run_profiled_code(

src/praisonai/praisonai/cli/commands/run.py

@@ -22,13 +22,14 @@ def run_main(
     framework: Optional[str] = typer.Option(None, "--framework", "-f", help="Framework: praisonai, crewai, autogen"),
     interactive: bool = typer.Option(False, "--interactive", "-i", help="Interactive mode"),
     verbose: bool = typer.Option(False, "--verbose", "-v", help="Verbose output"),
-    stream: bool = typer.Option(True, "--stream/--no-stream", help="Stream output"),
+    stream: bool = typer.Option(False, "--stream/--no-stream", help="Stream output (default: off for production use)"),
     trace: bool = typer.Option(False, "--trace", help="Enable tracing"),
     memory: bool = typer.Option(False, "--memory", help="Enable memory"),
     tools: Optional[str] = typer.Option(None, "--tools", "-t", help="Tools file path"),
     max_tokens: int = typer.Option(16000, "--max-tokens", help="Maximum output tokens"),
     profile: bool = typer.Option(False, "--profile", help="Enable CLI profiling (timing breakdown)"),
     profile_deep: bool = typer.Option(False, "--profile-deep", help="Enable deep profiling (cProfile stats, higher overhead)"),
+    output_mode: Optional[str] = typer.Option(None, "--output", "-o", help="Output mode: silent (default), actions, verbose, json, stream"),
 ):
     """
     Run agents from a file or prompt.
@@ -108,6 +109,7 @@ def run_main(
             memory=memory,
             tools=tools,
             max_tokens=max_tokens,
+            output_mode=output_mode,
         )
     else:
         # Run as prompt
@@ -120,6 +122,7 @@ def run_main(
             memory=memory,
             tools=tools,
             max_tokens=max_tokens,
+            output_mode=output_mode,
         )
 
 
@@ -134,6 +137,7 @@ def _run_from_file(
     memory: bool = False,
     tools: Optional[str] = None,
     max_tokens: int = 16000,
+    output_mode: Optional[str] = None,
 ):
     """Run agents from a YAML file."""
     output = get_output_controller()
@@ -178,12 +182,37 @@ def _run_prompt(
     memory: bool = False,
     tools: Optional[str] = None,
     max_tokens: int = 16000,
+    output_mode: Optional[str] = None,
 ):
     """Run a direct prompt."""
     output = get_output_controller()
     
     try:
-        # Use existing handle_direct_prompt
+        # If output_mode is "actions", use direct Agent with actions preset
+        if output_mode == "actions":
+            from praisonaiagents import Agent
+            
+            agent_config = {
+                "name": "RunAgent",
+                "role": "Assistant", 
+                "goal": "Complete the task",
+                "output": "actions",  # Use actions preset
+            }
+            if model:
+                agent_config["llm"] = model
+            
+            agent = Agent(**agent_config)
+            result = agent.start(prompt)
+            
+            output.emit_result(
+                message="Prompt completed",
+                data={"result": str(result) if result else None}
+            )
+            
+            # Don't print result again - actions mode already shows output
+            return
+        
+        # Use existing handle_direct_prompt for other modes
         from praisonai.cli.main import PraisonAI
         
         praison = PraisonAI()

src/praisonai/praisonai/cli/commands/test.py

@@ -10,6 +10,12 @@ Usage:
     praisonai test --provider openai    # Run OpenAI provider tests
     praisonai test --live               # Enable live API tests
     praisonai test --parallel auto      # Run with parallelization
+    
+    # Interactive mode tests
+    praisonai test interactive --csv tests.csv
+    praisonai test interactive --suite smoke
+    praisonai test interactive --list
+    praisonai test interactive --generate-template
 """
 
 import os
@@ -239,6 +245,272 @@ def run(
         raise typer.Exit(1)
 
 
+# Built-in interactive test suites
+BUILTIN_SUITES = {
+    "smoke": "smoke.csv",
+    "tools": "tools.csv",
+    "refactor": "refactor.csv",
+    "multi_agent": "multi_agent.csv",
+}
+
+# Special suites that don't use CSV
+SPECIAL_SUITES = {
+    "advanced": "Advanced Agent-Centric (5 complex autonomous scenarios)",
+}
+
+
+def _get_builtin_suite_path(suite: str):
+    """Get path to built-in test suite."""
+    from pathlib import Path
+    
+    if suite not in BUILTIN_SUITES:
+        return None
+    
+    # Find fixtures directory relative to package
+    package_dir = Path(__file__).parent.parent.parent.parent  # Up to src/praisonai
+    fixtures_dir = package_dir / "tests" / "fixtures" / "interactive"
+    
+    suite_path = fixtures_dir / BUILTIN_SUITES[suite]
+    if suite_path.exists():
+        return suite_path
+    
+    return None
+
+
+@app.command("interactive")
+def interactive(
+    csv: Optional[str] = typer.Option(
+        None, "--csv", "-c",
+        help="Path to CSV file with test cases"
+    ),
+    suite: Optional[str] = typer.Option(
+        None, "--suite", "-s",
+        help="Built-in test suite: smoke, tools, refactor, multi_agent"
+    ),
+    model: str = typer.Option(
+        "gpt-4o-mini", "--model", "-m",
+        help="LLM model for agent"
+    ),
+    judge_model: str = typer.Option(
+        "gpt-4o-mini", "--judge-model",
+        help="LLM model for judge"
+    ),
+    workspace: Optional[str] = typer.Option(
+        None, "--workspace", "-w",
+        help="Workspace directory (default: temp)"
+    ),
+    artifacts_dir: Optional[str] = typer.Option(
+        None, "--artifacts-dir",
+        help="Directory for test artifacts"
+    ),
+    fail_fast: bool = typer.Option(
+        False, "--fail-fast", "-x",
+        help="Stop on first failure"
+    ),
+    keep_artifacts: bool = typer.Option(
+        False, "--keep-artifacts",
+        help="Keep test artifacts after run"
+    ),
+    no_judge: bool = typer.Option(
+        False, "--no-judge",
+        help="Skip judge evaluation even if rubric present"
+    ),
+    verbose: bool = typer.Option(
+        False, "--verbose", "-v",
+        help="Verbose output"
+    ),
+    list_suites: bool = typer.Option(
+        False, "--list",
+        help="List available built-in suites"
+    ),
+    generate_template: bool = typer.Option(
+        False, "--generate-template",
+        help="Generate CSV template in current directory"
+    ),
+):
+    """
+    Run interactive mode tests from CSV.
+    
+    Tests are executed through the headless interactive core, using the same
+    pipeline as the interactive TUI (InteractiveRuntime, get_interactive_tools,
+    ActionOrchestrator).
+    
+    Examples:
+        praisonai test interactive --csv tests.csv
+        praisonai test interactive --suite smoke
+        praisonai test interactive --suite tools --keep-artifacts
+        praisonai test interactive --list
+        praisonai test interactive --generate-template
+    """
+    from pathlib import Path
+    
+    # List suites
+    if list_suites:
+        typer.echo("📋 Available built-in test suites:")
+        typer.echo("")
+        typer.echo("CSV-based suites:")
+        for name, filename in BUILTIN_SUITES.items():
+            suite_path = _get_builtin_suite_path(name)
+            status = "✅" if suite_path else "❌ (not found)"
+            typer.echo(f"  {name:15} {filename:20} {status}")
+        typer.echo("")
+        typer.echo("Special suites:")
+        for name, description in SPECIAL_SUITES.items():
+            typer.echo(f"  {name:15} {description}")
+        typer.echo("")
+        typer.echo("Usage: praisonai test interactive --suite <name>")
+        typer.echo("       PRAISONAI_LIVE_INTERACTIVE=1 praisonai test interactive --suite advanced")
+        return
+    
+    # Generate template
+    if generate_template:
+        # Lazy import
+        from praisonai.cli.features.csv_test_runner import generate_csv_template
+        
+        output_path = Path.cwd() / "interactive_tests_template.csv"
+        generate_csv_template(output_path)
+        typer.echo(f"✅ Generated CSV template: {output_path}")
+        return
+    
+    # Validate inputs
+    if not csv and not suite:
+        typer.echo("❌ Either --csv or --suite is required")
+        typer.echo("")
+        typer.echo("Examples:")
+        typer.echo("  praisonai test interactive --csv tests.csv")
+        typer.echo("  praisonai test interactive --suite smoke")
+        typer.echo("  praisonai test interactive --suite advanced")
+        typer.echo("  praisonai test interactive --list")
+        raise typer.Exit(1)
+    
+    # Handle special suites
+    if suite and suite in SPECIAL_SUITES:
+        if suite == "advanced":
+            _run_advanced_suite(
+                model=model,
+                artifacts_dir=Path(artifacts_dir) if artifacts_dir else None,
+                keep_artifacts=keep_artifacts,
+                verbose=verbose,
+            )
+            return
+    
+    # Get CSV path
+    if csv:
+        csv_path = Path(csv)
+        if not csv_path.exists():
+            typer.echo(f"❌ CSV file not found: {csv}")
+            raise typer.Exit(1)
+    else:
+        csv_path = _get_builtin_suite_path(suite)
+        if not csv_path:
+            typer.echo(f"❌ Built-in suite not found: {suite}")
+            typer.echo("Use --list to see available suites")
+            raise typer.Exit(1)
+    
+    # Lazy import runner
+    from praisonai.cli.features.csv_test_runner import CSVTestRunner
+    
+    typer.echo(f"🧪 Running interactive tests from: {csv_path}")
+    typer.echo(f"   Model: {model} | Judge: {judge_model}")
+    typer.echo("")
+    
+    # Create runner
+    runner = CSVTestRunner(
+        csv_path=csv_path,
+        model=model,
+        judge_model=judge_model,
+        workspace=Path(workspace) if workspace else None,
+        artifacts_dir=Path(artifacts_dir) if artifacts_dir else None,
+        fail_fast=fail_fast,
+        keep_artifacts=keep_artifacts,
+        no_judge=no_judge,
+        verbose=verbose,
+    )
+    
+    # Run tests
+    summary = runner.run()
+    
+    # Print summary
+    typer.echo("")
+    summary.print_summary()
+    
+    # Save summary if keeping artifacts
+    if keep_artifacts and runner.artifacts_dir:
+        import json
+        summary_path = runner.artifacts_dir / "summary.json"
+        summary_path.parent.mkdir(parents=True, exist_ok=True)
+        with open(summary_path, "w") as f:
+            json.dump(summary.to_dict(), f, indent=2, default=str)
+        typer.echo(f"\n📁 Artifacts saved to: {runner.artifacts_dir}")
+    
+    # Exit code based on results
+    if summary.failed > 0 or summary.errors > 0:
+        raise typer.Exit(1)
+
+
+def _run_advanced_suite(
+    model: str = "gpt-4o-mini",
+    artifacts_dir = None,
+    keep_artifacts: bool = True,
+    verbose: bool = True,
+):
+    """Run the Advanced Agent-Centric test suite."""
+    from pathlib import Path
+    
+    # Check prerequisites
+    typer.echo("🔍 Checking Advanced suite prerequisites...")
+    
+    # Check if live interactive is enabled
+    live_enabled = os.environ.get("PRAISONAI_LIVE_INTERACTIVE", "0") == "1"
+    has_api_key = bool(os.environ.get("OPENAI_API_KEY"))
+    
+    if not live_enabled:
+        typer.echo("\n⚠️  Cannot run Advanced tests: PRAISONAI_LIVE_INTERACTIVE=1 not set")
+        typer.echo("")
+        typer.echo("Prerequisites:")
+        typer.echo("  1. Set PRAISONAI_LIVE_INTERACTIVE=1")
+        typer.echo("  2. Set OPENAI_API_KEY")
+        typer.echo("")
+        typer.echo("Example:")
+        typer.echo("  PRAISONAI_LIVE_INTERACTIVE=1 praisonai test interactive --suite advanced")
+        raise typer.Exit(1)
+    
+    if not has_api_key:
+        typer.echo("\n⚠️  Cannot run Advanced tests: OPENAI_API_KEY not set")
+        raise typer.Exit(1)
+    
+    typer.echo("✅ Live interactive mode enabled")
+    typer.echo("✅ API key configured")
+    typer.echo("")
+    
+    # Lazy import runner and scenarios
+    from tests.live.interactive.runner import LiveInteractiveRunner
+    from tests.live.interactive.advanced.scenarios import ALL_ADVANCED_SCENARIOS
+    
+    # Create runner
+    runner = LiveInteractiveRunner(
+        model=model,
+        artifacts_dir=Path(artifacts_dir) if artifacts_dir else None,
+        keep_artifacts=keep_artifacts,
+        verbose=verbose,
+    )
+    
+    # Run tests
+    summary = runner.run_all(ALL_ADVANCED_SCENARIOS)
+    
+    # Print summary
+    typer.echo("")
+    summary.print_summary()
+    
+    # Show artifacts location
+    if keep_artifacts:
+        typer.echo(f"\n📁 Artifacts saved to: {runner.artifacts_dir}")
+    
+    # Exit code
+    if summary.failed > 0:
+        raise typer.Exit(1)
+
+
 @app.command()
 def info():
     """Show test configuration and available options."""

src/praisonai/praisonai/cli/features/action_orchestrator.py

@@ -160,6 +160,8 @@ class ActionOrchestrator:
     
     async def _create_fallback_plan(self, prompt: str, plan_id: str) -> ActionResult:
         """Create a fallback plan when ACP is not available for planning."""
+        import re  # Import at function level for all pattern matching
+        
         # This is a simplified fallback - in production, this would use LLM
         steps = []
         
@@ -168,7 +170,6 @@ class ActionOrchestrator:
         
         if "create" in prompt_lower or "new file" in prompt_lower:
             # Extract file path if mentioned
-            import re
             file_match = re.search(r'(?:create|new)\s+(?:a\s+)?(?:file\s+)?(?:called\s+)?([^\s]+\.\w+)', prompt_lower)
             target = file_match.group(1) if file_match else "new_file.py"
             
@@ -188,7 +189,8 @@ class ActionOrchestrator:
             ))
         
         elif "edit" in prompt_lower or "modify" in prompt_lower or "change" in prompt_lower:
-            file_match = re.search(r'(?:edit|modify|change)\s+([^\s]+\.\w+)', prompt_lower)
+            # Match patterns like "edit file X", "edit X", "modify X"
+            file_match = re.search(r'(?:edit|modify|change)\s+(?:file\s+)?([^\s]+(?:\.\w+)?)', prompt_lower)
             target = file_match.group(1) if file_match else ""
             
             if target:
@@ -197,11 +199,13 @@ class ActionOrchestrator:
                     action_type=ActionType.FILE_EDIT,
                     description=f"Edit file: {target}",
                     target=target,
-                    params={"changes": prompt}
+                    params={"changes": prompt},
+                    status=ActionStatus.APPROVED  # Auto-approve for fallback
                 ))
         
         elif "delete" in prompt_lower or "remove" in prompt_lower:
-            file_match = re.search(r'(?:delete|remove)\s+([^\s]+\.\w+)', prompt_lower)
+            # Match patterns like "delete file X", "delete X", "remove X"
+            file_match = re.search(r'(?:delete|remove)\s+(?:file\s+)?([^\s]+(?:\.\w+)?)', prompt_lower)
             target = file_match.group(1) if file_match else ""
             
             if target:
@@ -209,7 +213,8 @@ class ActionOrchestrator:
                     id=self._generate_step_id(plan_id, 0),
                     action_type=ActionType.FILE_DELETE,
                     description=f"Delete file: {target}",
-                    target=target
+                    target=target,
+                    status=ActionStatus.APPROVED  # Auto-approve for fallback
                 ))
         
         elif "rename" in prompt_lower:
@@ -218,19 +223,22 @@ class ActionOrchestrator:
                 action_type=ActionType.REFACTOR,
                 description="Rename operation",
                 target="",
-                params={"prompt": prompt}
+                params={"prompt": prompt},
+                status=ActionStatus.APPROVED  # Auto-approve for fallback
             ))
         
-        elif "run" in prompt_lower or "execute" in prompt_lower:
-            cmd_match = re.search(r'(?:run|execute)\s+[`"\']?([^`"\']+)[`"\']?', prompt_lower)
-            command = cmd_match.group(1) if cmd_match else ""
+        elif "run" in prompt_lower or "execute" in prompt_lower or "command" in prompt_lower:
+            # Match patterns like "run command: X", "execute X", "run X"
+            cmd_match = re.search(r'(?:run|execute)\s+(?:command[:\s]+)?[`"\']?(.+?)[`"\']?$', prompt_lower)
+            command = cmd_match.group(1).strip() if cmd_match else ""
             
             if command:
                 steps.append(ActionStep(
                     id=self._generate_step_id(plan_id, 0),
                     action_type=ActionType.SHELL_COMMAND,
                     description=f"Execute: {command}",
-                    target=command
+                    target=command,
+                    status=ActionStatus.APPROVED  # Auto-approve for fallback
                 ))
         
         if not steps:

src/praisonai/praisonai/cli/features/code_intelligence.py

@@ -596,20 +596,34 @@ class CodeIntelligenceRouter:
     
     def _extract_symbol_from_query(self, query: str) -> Optional[str]:
         """Extract symbol name from query."""
-        # Common patterns for symbol extraction
+        # Common patterns for symbol extraction (order matters - more specific first)
         patterns = [
-            r'(?:definition\s+of|where\s+is|find|show)\s+["\']?(\w+)["\']?',
+            # "find all references to X" - capture X after "to"
+            r'references?\s+to\s+["\']?(\w+)["\']?',
+            # "definition of X" - capture X after "of"
+            r'definition\s+of\s+["\']?(\w+)["\']?',
+            # "where is X defined/used"
+            r'where\s+is\s+["\']?(\w+)["\']?\s+(?:defined|used)',
+            # Quoted or backticked symbols
             r'["\'](\w+)["\']',
             r'`(\w+)`',
+            # "X is defined/used"
             r'(\w+)\s+(?:is\s+)?(?:defined|used)',
+            # Generic "find/show X" - but filter common words
+            r'(?:find|show)\s+["\']?(\w+)["\']?',
         ]
         
+        # Words to filter out
+        filter_words = {'the', 'a', 'an', 'is', 'are', 'in', 'of', 'to', 'for', 
+                       'all', 'any', 'some', 'this', 'that', 'where', 'what',
+                       'definition', 'references', 'reference', 'symbol', 'symbols'}
+        
         for pattern in patterns:
             match = re.search(pattern, query, re.IGNORECASE)
             if match:
                 symbol = match.group(1)
                 # Filter out common words
-                if symbol.lower() not in ['the', 'a', 'an', 'is', 'are', 'in', 'of', 'to', 'for']:
+                if symbol.lower() not in filter_words:
                     return symbol
         
         return None

src/praisonai/praisonai/cli/features/csv_test_runner.py

@@ -0,0 +1,736 @@
+"""
+CSV-driven Interactive Test Runner for PraisonAI.
+
+Provides a CSV-driven test runner that:
+- Loads test cases from CSV with defined schema
+- Runs each test in isolated temp workspace
+- Executes prompts via headless interactive core
+- Validates tool calls, files, and responses
+- Integrates LLM-as-judge for response quality
+- Emits artifacts per-test
+"""
+
+import csv
+import json
+import logging
+import os
+import time
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+logger = logging.getLogger(__name__)
+
+
+# CSV Schema definition
+CSV_SCHEMA = {
+    # Required columns
+    "id": {"required": True, "type": str, "description": "Unique test identifier"},
+    "name": {"required": True, "type": str, "description": "Test name"},
+    "prompts": {"required": True, "type": str, "description": "Single prompt or JSON array"},
+    # Optional columns
+    "description": {"required": False, "type": str, "default": ""},
+    "mode": {"required": False, "type": str, "default": "headless"},
+    "workspace_fixture": {"required": False, "type": str, "default": "empty"},
+    "expected_tools": {"required": False, "type": str, "default": ""},
+    "forbidden_tools": {"required": False, "type": str, "default": ""},
+    "expected_files": {"required": False, "type": str, "default": "{}"},
+    "expected_response": {"required": False, "type": str, "default": ""},
+    "judge_rubric": {"required": False, "type": str, "default": ""},
+    "judge_threshold": {"required": False, "type": float, "default": 7.0},
+    "judge_model": {"required": False, "type": str, "default": "gpt-4o-mini"},
+    "timeout": {"required": False, "type": int, "default": 60},
+    "retries": {"required": False, "type": int, "default": 0},
+    "skip_if": {"required": False, "type": str, "default": ""},
+    "agents": {"required": False, "type": str, "default": "[]"},
+    "workflow": {"required": False, "type": str, "default": "{}"},
+}
+
+
+@dataclass
+class TestCase:
+    """A single test case from CSV."""
+    id: str
+    name: str
+    prompts: List[str]
+    description: str = ""
+    mode: str = "headless"
+    workspace_fixture: str = "empty"
+    expected_tools: List[str] = field(default_factory=list)
+    forbidden_tools: List[str] = field(default_factory=list)
+    expected_files: Dict[str, str] = field(default_factory=dict)
+    expected_response: str = ""
+    judge_rubric: str = ""
+    judge_threshold: float = 7.0
+    judge_model: str = "gpt-4o-mini"
+    timeout: int = 60
+    retries: int = 0
+    skip_if: str = ""
+    agents: List[Dict[str, Any]] = field(default_factory=list)
+    workflow: Dict[str, Any] = field(default_factory=dict)
+    
+    def should_skip(self) -> Optional[str]:
+        """Check if test should be skipped."""
+        if not self.skip_if:
+            return None
+        
+        conditions = [c.strip() for c in self.skip_if.split(",")]
+        
+        for condition in conditions:
+            if condition == "no_openai_key":
+                if not os.environ.get("OPENAI_API_KEY"):
+                    return "OPENAI_API_KEY not set"
+            elif condition == "no_anthropic_key":
+                if not os.environ.get("ANTHROPIC_API_KEY"):
+                    return "ANTHROPIC_API_KEY not set"
+            elif condition == "no_lsp":
+                # Could check if LSP is available
+                pass
+            elif condition == "no_network":
+                if os.environ.get("PRAISONAI_NO_NETWORK"):
+                    return "Network disabled"
+        
+        return None
+
+
+@dataclass
+class TestResult:
+    """Result of a single test."""
+    test_id: str
+    test_name: str
+    status: str  # passed, failed, skipped, error
+    duration: float
+    tool_calls: List[str] = field(default_factory=list)
+    response: str = ""
+    judge_score: Optional[float] = None
+    judge_passed: Optional[bool] = None
+    judge_reasoning: str = ""
+    artifacts_path: Optional[str] = None
+    error: Optional[str] = None
+    skip_reason: Optional[str] = None
+    assertions: Dict[str, bool] = field(default_factory=dict)
+    
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "test_id": self.test_id,
+            "test_name": self.test_name,
+            "status": self.status,
+            "duration": self.duration,
+            "tool_calls": self.tool_calls,
+            "response": self.response[:500] if self.response else "",
+            "judge_score": self.judge_score,
+            "judge_passed": self.judge_passed,
+            "judge_reasoning": self.judge_reasoning,
+            "artifacts_path": self.artifacts_path,
+            "error": self.error,
+            "skip_reason": self.skip_reason,
+            "assertions": self.assertions,
+        }
+
+
+@dataclass
+class TestSummary:
+    """Summary of all test results."""
+    total: int = 0
+    passed: int = 0
+    failed: int = 0
+    skipped: int = 0
+    errors: int = 0
+    duration: float = 0.0
+    judge_avg: Optional[float] = None
+    results: List[TestResult] = field(default_factory=list)
+    
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "total": self.total,
+            "passed": self.passed,
+            "failed": self.failed,
+            "skipped": self.skipped,
+            "errors": self.errors,
+            "duration": self.duration,
+            "judge_avg": self.judge_avg,
+            "pass_rate": self.passed / self.total if self.total > 0 else 0,
+            "results": [r.to_dict() for r in self.results],
+        }
+    
+    def print_summary(self) -> None:
+        """Print summary to console."""
+        try:
+            from rich.console import Console
+            from rich.table import Table
+            
+            console = Console()
+            
+            table = Table(title="Interactive Test Results")
+            table.add_column("ID", style="cyan")
+            table.add_column("Name", style="white")
+            table.add_column("Status")
+            table.add_column("Duration", justify="right")
+            table.add_column("Judge", justify="right")
+            
+            for r in self.results:
+                status_style = {
+                    "passed": "[green]✓ passed[/green]",
+                    "failed": "[red]✗ failed[/red]",
+                    "skipped": "[yellow]○ skipped[/yellow]",
+                    "error": "[red]! error[/red]",
+                }.get(r.status, r.status)
+                
+                judge_str = f"{r.judge_score:.1f}" if r.judge_score is not None else "-"
+                
+                table.add_row(
+                    r.test_id,
+                    r.test_name[:30],
+                    status_style,
+                    f"{r.duration:.2f}s",
+                    judge_str,
+                )
+            
+            console.print(table)
+            console.print()
+            console.print(f"[bold]Total:[/bold] {self.total} | "
+                         f"[green]Passed:[/green] {self.passed} | "
+                         f"[red]Failed:[/red] {self.failed} | "
+                         f"[yellow]Skipped:[/yellow] {self.skipped}")
+            if self.judge_avg is not None:
+                console.print(f"[bold]Average Judge Score:[/bold] {self.judge_avg:.2f}")
+            console.print(f"[bold]Total Duration:[/bold] {self.duration:.2f}s")
+            
+        except ImportError:
+            # Fallback to plain print
+            print("\n=== Interactive Test Results ===")
+            for r in self.results:
+                status = {"passed": "✓", "failed": "✗", "skipped": "○", "error": "!"}.get(r.status, "?")
+                print(f"  {status} {r.test_id}: {r.test_name} ({r.duration:.2f}s)")
+            print(f"\nTotal: {self.total} | Passed: {self.passed} | Failed: {self.failed} | Skipped: {self.skipped}")
+
+
+def parse_csv(csv_path: Path) -> List[TestCase]:
+    """
+    Parse CSV file into TestCase objects.
+    
+    Args:
+        csv_path: Path to CSV file
+        
+    Returns:
+        List of TestCase objects
+    """
+    test_cases = []
+    
+    with open(csv_path, "r", encoding="utf-8") as f:
+        reader = csv.DictReader(f)
+        
+        for row_num, row in enumerate(reader, start=2):  # Start at 2 (header is 1)
+            try:
+                # Parse prompts (single string or JSON array)
+                prompts_raw = row.get("prompts", "").strip()
+                if prompts_raw.startswith("["):
+                    prompts = json.loads(prompts_raw)
+                else:
+                    prompts = [prompts_raw] if prompts_raw else []
+                
+                # Parse expected_tools
+                expected_tools = []
+                if row.get("expected_tools"):
+                    expected_tools = [t.strip() for t in row["expected_tools"].split(",") if t.strip()]
+                
+                # Parse forbidden_tools
+                forbidden_tools = []
+                if row.get("forbidden_tools"):
+                    forbidden_tools = [t.strip() for t in row["forbidden_tools"].split(",") if t.strip()]
+                
+                # Parse expected_files JSON
+                expected_files = {}
+                if row.get("expected_files"):
+                    try:
+                        expected_files = json.loads(row["expected_files"])
+                    except json.JSONDecodeError:
+                        logger.warning(f"Row {row_num}: Invalid JSON in expected_files")
+                
+                # Parse agents JSON
+                agents = []
+                if row.get("agents"):
+                    try:
+                        agents = json.loads(row["agents"])
+                    except json.JSONDecodeError:
+                        logger.warning(f"Row {row_num}: Invalid JSON in agents")
+                
+                # Parse workflow JSON
+                workflow = {}
+                if row.get("workflow"):
+                    try:
+                        workflow = json.loads(row["workflow"])
+                    except json.JSONDecodeError:
+                        logger.warning(f"Row {row_num}: Invalid JSON in workflow")
+                
+                # Parse numeric fields
+                judge_threshold = float(row.get("judge_threshold", 7.0) or 7.0)
+                timeout = int(row.get("timeout", 60) or 60)
+                retries = int(row.get("retries", 0) or 0)
+                
+                test_case = TestCase(
+                    id=row.get("id", f"test_{row_num}"),
+                    name=row.get("name", f"Test {row_num}"),
+                    prompts=prompts,
+                    description=row.get("description", ""),
+                    mode=row.get("mode", "headless"),
+                    workspace_fixture=row.get("workspace_fixture", "empty"),
+                    expected_tools=expected_tools,
+                    forbidden_tools=forbidden_tools,
+                    expected_files=expected_files,
+                    expected_response=row.get("expected_response", ""),
+                    judge_rubric=row.get("judge_rubric", ""),
+                    judge_threshold=judge_threshold,
+                    judge_model=row.get("judge_model", "gpt-4o-mini"),
+                    timeout=timeout,
+                    retries=retries,
+                    skip_if=row.get("skip_if", ""),
+                    agents=agents,
+                    workflow=workflow,
+                )
+                
+                test_cases.append(test_case)
+                
+            except Exception as e:
+                logger.error(f"Error parsing row {row_num}: {e}")
+                continue
+    
+    return test_cases
+
+
+class CSVTestRunner:
+    """
+    CSV-driven test runner for interactive mode.
+    
+    Usage:
+        runner = CSVTestRunner(csv_path="tests.csv")
+        summary = runner.run()
+        summary.print_summary()
+    """
+    
+    def __init__(
+        self,
+        csv_path: Path,
+        model: str = "gpt-4o-mini",
+        judge_model: str = "gpt-4o-mini",
+        workspace: Optional[Path] = None,
+        artifacts_dir: Optional[Path] = None,
+        fail_fast: bool = False,
+        keep_artifacts: bool = False,
+        no_judge: bool = False,
+        verbose: bool = False,
+    ):
+        """
+        Initialize CSV test runner.
+        
+        Args:
+            csv_path: Path to CSV file with test cases
+            model: LLM model for agent
+            judge_model: LLM model for judge
+            workspace: Base workspace directory
+            artifacts_dir: Directory for artifacts
+            fail_fast: Stop on first failure
+            keep_artifacts: Keep artifacts after run
+            no_judge: Skip judge even if rubric present
+            verbose: Verbose output
+        """
+        self.csv_path = Path(csv_path)
+        self.model = model
+        self.judge_model = judge_model
+        self.workspace = workspace
+        self.artifacts_dir = artifacts_dir
+        self.fail_fast = fail_fast
+        self.keep_artifacts = keep_artifacts
+        self.no_judge = no_judge
+        self.verbose = verbose
+        self.results: List[TestResult] = []
+    
+    def run(self) -> TestSummary:
+        """
+        Run all tests from CSV.
+        
+        Returns:
+            TestSummary with all results
+        """
+        start_time = time.time()
+        
+        # Parse CSV
+        test_cases = parse_csv(self.csv_path)
+        logger.info(f"Loaded {len(test_cases)} test cases from {self.csv_path}")
+        
+        if self.verbose:
+            print(f"Running {len(test_cases)} tests from {self.csv_path}")
+        
+        # Run each test
+        for i, test_case in enumerate(test_cases):
+            if self.verbose:
+                print(f"\n[{i+1}/{len(test_cases)}] Running: {test_case.name}")
+            
+            # Check skip conditions
+            skip_reason = test_case.should_skip()
+            if skip_reason:
+                result = TestResult(
+                    test_id=test_case.id,
+                    test_name=test_case.name,
+                    status="skipped",
+                    duration=0.0,
+                    skip_reason=skip_reason,
+                )
+                self.results.append(result)
+                if self.verbose:
+                    print(f"  Skipped: {skip_reason}")
+                continue
+            
+            # Run test
+            result = self._run_single_test(test_case)
+            self.results.append(result)
+            
+            if self.verbose:
+                status_icon = {"passed": "✓", "failed": "✗", "error": "!"}.get(result.status, "?")
+                print(f"  {status_icon} {result.status} ({result.duration:.2f}s)")
+                if result.error:
+                    print(f"    Error: {result.error}")
+            
+            # Fail fast
+            if self.fail_fast and result.status in ("failed", "error"):
+                logger.info(f"Stopping due to fail-fast: {test_case.id}")
+                break
+        
+        # Create summary
+        summary = self._create_summary(time.time() - start_time)
+        
+        return summary
+    
+    def _run_single_test(self, test_case: TestCase) -> TestResult:
+        """Run a single test case."""
+        import tempfile
+        
+        start_time = time.time()
+        
+        # Create workspace
+        if self.workspace:
+            test_workspace = self.workspace / test_case.id
+        else:
+            test_workspace = Path(tempfile.mkdtemp(prefix=f"praison_test_{test_case.id}_"))
+        
+        # Create artifacts dir
+        if self.artifacts_dir:
+            test_artifacts = self.artifacts_dir / test_case.id
+        else:
+            test_artifacts = test_workspace / "artifacts"
+        
+        try:
+            # Import harness (lazy)
+            from .interactive_test_harness import InteractiveTestHarness
+            
+            # Create harness
+            harness = InteractiveTestHarness(
+                workspace=test_workspace,
+                artifacts_dir=test_artifacts,
+                keep_workspace=self.keep_artifacts,
+            )
+            
+            # Setup workspace
+            harness.setup_workspace(test_case.workspace_fixture)
+            
+            # Run prompts
+            exec_result = harness.run(
+                prompts=test_case.prompts,
+                model=self.model,
+                approval_mode="auto",
+                agents=test_case.agents if test_case.agents else None,
+                workflow=test_case.workflow if test_case.workflow else None,
+            )
+            
+            duration = time.time() - start_time
+            
+            # Get response
+            response = "\n".join(exec_result.responses) if exec_result.responses else ""
+            
+            # Verify assertions
+            assertions = {}
+            
+            # Tool call assertions
+            tool_result = harness.verify_tool_calls(
+                expected_tools=test_case.expected_tools,
+                forbidden_tools=test_case.forbidden_tools,
+            )
+            assertions["tools"] = tool_result["passed"]
+            
+            # File assertions
+            if test_case.expected_files:
+                file_results = harness.verify_files(test_case.expected_files)
+                assertions["files"] = all(file_results.values())
+            else:
+                assertions["files"] = True
+            
+            # Response assertion
+            if test_case.expected_response:
+                assertions["response"] = harness.verify_response(response, test_case.expected_response)
+            else:
+                assertions["response"] = True
+            
+            # Judge evaluation
+            judge_score = None
+            judge_passed = None
+            judge_reasoning = ""
+            
+            if test_case.judge_rubric and not self.no_judge:
+                judge_result = self._run_judge(
+                    response=response,
+                    rubric=test_case.judge_rubric,
+                    threshold=test_case.judge_threshold,
+                    model=test_case.judge_model or self.judge_model,
+                )
+                judge_score = judge_result.get("score")
+                judge_passed = judge_result.get("passed")
+                judge_reasoning = judge_result.get("reasoning", "")
+                assertions["judge"] = judge_passed if judge_passed is not None else True
+                harness.artifacts.judge_result = judge_result
+            
+            # Determine overall status
+            all_passed = all(assertions.values()) and exec_result.success
+            status = "passed" if all_passed else "failed"
+            
+            # Snapshot workspace if keeping artifacts
+            if self.keep_artifacts:
+                harness.snapshot_workspace()
+            
+            # Save artifacts
+            artifacts_path = str(harness.save_artifacts(test_case.id))
+            
+            return TestResult(
+                test_id=test_case.id,
+                test_name=test_case.name,
+                status=status,
+                duration=duration,
+                tool_calls=harness._executor.get_tools_called() if harness._executor else [],
+                response=response,
+                judge_score=judge_score,
+                judge_passed=judge_passed,
+                judge_reasoning=judge_reasoning,
+                artifacts_path=artifacts_path,
+                assertions=assertions,
+            )
+            
+        except Exception as e:
+            logger.error(f"Test {test_case.id} error: {e}", exc_info=True)
+            return TestResult(
+                test_id=test_case.id,
+                test_name=test_case.name,
+                status="error",
+                duration=time.time() - start_time,
+                error=str(e),
+            )
+        finally:
+            if not self.keep_artifacts:
+                try:
+                    import shutil
+                    if test_workspace.exists() and not self.workspace:
+                        shutil.rmtree(test_workspace)
+                except Exception:
+                    pass
+    
+    def _run_judge(
+        self,
+        response: str,
+        rubric: str,
+        threshold: float,
+        model: str,
+    ) -> Dict[str, Any]:
+        """
+        Run LLM judge on response using Agent class.
+        
+        Args:
+            response: Response to evaluate
+            rubric: Evaluation rubric
+            threshold: Pass threshold (1-10)
+            model: Judge model
+            
+        Returns:
+            Dict with score, passed, reasoning
+        """
+        try:
+            from praisonaiagents import Agent
+            
+            # Create judge agent
+            judge_agent = Agent(
+                name="Judge",
+                role="Evaluator",
+                goal="Evaluate response quality based on rubric",
+                instructions=f"""You are an expert evaluator. Your task is to evaluate a response based on the given rubric.
+
+RUBRIC:
+{rubric}
+
+RESPONSE TO EVALUATE:
+{response}
+
+Provide your evaluation in the following format:
+SCORE: [1-10]
+REASONING: [Your detailed reasoning]
+
+Be strict but fair. A score of 7+ means the response adequately meets the rubric criteria.""",
+                llm=model,
+                verbose=False,
+            )
+            
+            # Get evaluation
+            eval_response = judge_agent.chat("Please evaluate the response based on the rubric.")
+            
+            # Parse score from response
+            score = self._parse_judge_score(eval_response)
+            passed = score >= threshold if score is not None else None
+            
+            return {
+                "score": score,
+                "passed": passed,
+                "reasoning": eval_response,
+                "threshold": threshold,
+                "model": model,
+            }
+            
+        except Exception as e:
+            logger.warning(f"Judge evaluation failed: {e}")
+            return {
+                "score": None,
+                "passed": None,
+                "reasoning": f"Judge error: {e}",
+                "error": str(e),
+            }
+    
+    def _parse_judge_score(self, response: str) -> Optional[float]:
+        """Parse score from judge response."""
+        import re
+        
+        # Try to find "SCORE: X" pattern
+        match = re.search(r"SCORE:\s*(\d+(?:\.\d+)?)", response, re.IGNORECASE)
+        if match:
+            try:
+                return float(match.group(1))
+            except ValueError:
+                pass
+        
+        # Try to find standalone number at start
+        match = re.search(r"^(\d+(?:\.\d+)?)", response.strip())
+        if match:
+            try:
+                score = float(match.group(1))
+                if 1 <= score <= 10:
+                    return score
+            except ValueError:
+                pass
+        
+        return None
+    
+    def _create_summary(self, total_duration: float) -> TestSummary:
+        """Create test summary."""
+        summary = TestSummary(
+            total=len(self.results),
+            passed=sum(1 for r in self.results if r.status == "passed"),
+            failed=sum(1 for r in self.results if r.status == "failed"),
+            skipped=sum(1 for r in self.results if r.status == "skipped"),
+            errors=sum(1 for r in self.results if r.status == "error"),
+            duration=total_duration,
+            results=self.results,
+        )
+        
+        # Calculate average judge score
+        judge_scores = [r.judge_score for r in self.results if r.judge_score is not None]
+        if judge_scores:
+            summary.judge_avg = sum(judge_scores) / len(judge_scores)
+        
+        return summary
+
+
+def generate_csv_template(output_path: Path = None) -> str:
+    """
+    Generate a CSV template with all columns.
+    
+    Args:
+        output_path: Path to write template (optional)
+        
+    Returns:
+        CSV template string
+    """
+    headers = list(CSV_SCHEMA.keys())
+    
+    # Example rows
+    examples = [
+        {
+            "id": "smoke_01",
+            "name": "Basic Chat",
+            "prompts": "Hello, what is 2+2?",
+            "description": "Test basic chat response",
+            "mode": "headless",
+            "workspace_fixture": "empty",
+            "expected_tools": "",
+            "forbidden_tools": "",
+            "expected_files": "{}",
+            "expected_response": "4",
+            "judge_rubric": "Response contains the number 4",
+            "judge_threshold": "7.0",
+            "judge_model": "gpt-4o-mini",
+            "timeout": "60",
+            "retries": "0",
+            "skip_if": "",
+            "agents": "[]",
+            "workflow": "{}",
+        },
+        {
+            "id": "tools_01",
+            "name": "Create File",
+            "prompts": "Create a file called hello.py with print('hello')",
+            "description": "Test file creation",
+            "mode": "headless",
+            "workspace_fixture": "empty",
+            "expected_tools": "acp_create_file",
+            "forbidden_tools": "",
+            "expected_files": '{"hello.py": "print.*hello"}',
+            "expected_response": "",
+            "judge_rubric": "File was created successfully",
+            "judge_threshold": "7.0",
+            "judge_model": "gpt-4o-mini",
+            "timeout": "60",
+            "retries": "0",
+            "skip_if": "no_openai_key",
+            "agents": "[]",
+            "workflow": "{}",
+        },
+        {
+            "id": "multi_01",
+            "name": "Multi-step Edit",
+            "prompts": '["Create test.py with x=1", "Edit test.py to change x=1 to x=2"]',
+            "description": "Test multi-step editing",
+            "mode": "headless",
+            "workspace_fixture": "empty",
+            "expected_tools": "acp_create_file,acp_edit_file",
+            "forbidden_tools": "",
+            "expected_files": '{"test.py": "x.*=.*2"}',
+            "expected_response": "",
+            "judge_rubric": "",
+            "judge_threshold": "7.0",
+            "judge_model": "gpt-4o-mini",
+            "timeout": "120",
+            "retries": "0",
+            "skip_if": "no_openai_key",
+            "agents": "[]",
+            "workflow": "{}",
+        },
+    ]
+    
+    # Build CSV string
+    import io
+    output = io.StringIO()
+    writer = csv.DictWriter(output, fieldnames=headers)
+    writer.writeheader()
+    for example in examples:
+        writer.writerow(example)
+    
+    csv_content = output.getvalue()
+    
+    if output_path:
+        Path(output_path).write_text(csv_content)
+        logger.info(f"Generated CSV template at {output_path}")
+    
+    return csv_content

src/praisonai/praisonai/cli/features/interactive_core_headless.py

@@ -0,0 +1,450 @@
+"""
+Headless Interactive Core Executor for PraisonAI.
+
+This module provides a headless execution mode that uses the SAME interactive core
+pipeline as the interactive TUI:
+- InteractiveRuntime config
+- get_interactive_tools() tool resolution
+- ActionOrchestrator behavior (plan/approve/apply/verify)
+- Tool call tracing
+
+This is for tests and CI - no terminal/prompt_toolkit dependency.
+"""
+
+import functools
+import logging
+import os
+import time
+from dataclasses import dataclass, field
+from datetime import datetime
+from typing import Any, Callable, Dict, List, Optional, Tuple, Union
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class ToolCallTrace:
+    """Record of a single tool call."""
+    tool_name: str
+    args: Tuple
+    kwargs: Dict[str, Any]
+    result: Any
+    success: bool
+    error: Optional[str]
+    duration: float
+    timestamp: str
+    
+    def to_dict(self) -> Dict[str, Any]:
+        """Convert to dictionary for JSON serialization."""
+        return {
+            "tool_name": self.tool_name,
+            "args": _redact_sensitive(self.args),
+            "kwargs": _redact_sensitive(self.kwargs),
+            "result": _truncate(str(self.result), 500) if self.result else None,
+            "success": self.success,
+            "error": self.error,
+            "duration": self.duration,
+            "timestamp": self.timestamp,
+        }
+
+
+@dataclass
+class HeadlessExecutionResult:
+    """Result of a headless execution."""
+    success: bool
+    responses: List[str]
+    tool_trace: List[ToolCallTrace]
+    duration: float
+    error: Optional[str] = None
+    transcript: List[Dict[str, str]] = field(default_factory=list)
+    
+    def to_dict(self) -> Dict[str, Any]:
+        """Convert to dictionary."""
+        return {
+            "success": self.success,
+            "responses": self.responses,
+            "tool_trace": [t.to_dict() for t in self.tool_trace],
+            "duration": self.duration,
+            "error": self.error,
+            "transcript": self.transcript,
+        }
+
+
+@dataclass
+class AgentConfig:
+    """Configuration for a single agent in headless mode."""
+    name: str = "HeadlessAgent"
+    instructions: str = "You are a helpful assistant with file and code tools."
+    role: str = "assistant"
+    llm: str = "gpt-4o-mini"
+    
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "name": self.name,
+            "instructions": self.instructions,
+            "role": self.role,
+            "llm": self.llm,
+        }
+
+
+@dataclass 
+class HeadlessConfig:
+    """Configuration for headless interactive core execution."""
+    workspace: str = field(default_factory=os.getcwd)
+    model: str = "gpt-4o-mini"
+    approval_mode: str = "auto"  # auto, manual, scoped
+    enable_acp: bool = True
+    enable_lsp: bool = True
+    enable_basic: bool = True
+    timeout: int = 60
+    verbose: bool = False
+    # Multi-agent support
+    agents: List[AgentConfig] = field(default_factory=list)
+    workflow: Optional[Dict[str, Any]] = None  # Turn routing/handoffs
+    
+    def __post_init__(self):
+        if not self.agents:
+            # Default single agent
+            self.agents = [AgentConfig(llm=self.model)]
+
+
+def _redact_sensitive(obj: Any, sensitive_keys: set = None) -> Any:
+    """Redact sensitive information from args/kwargs."""
+    if sensitive_keys is None:
+        sensitive_keys = {"api_key", "password", "secret", "token", "key"}
+    
+    if isinstance(obj, dict):
+        return {
+            k: "[REDACTED]" if any(s in k.lower() for s in sensitive_keys) else _redact_sensitive(v, sensitive_keys)
+            for k, v in obj.items()
+        }
+    elif isinstance(obj, (list, tuple)):
+        return type(obj)(_redact_sensitive(item, sensitive_keys) for item in obj)
+    elif isinstance(obj, str) and len(obj) > 100:
+        return obj[:100] + "...[truncated]"
+    return obj
+
+
+def _truncate(s: str, max_len: int) -> str:
+    """Truncate string to max length."""
+    if len(s) <= max_len:
+        return s
+    return s[:max_len] + "...[truncated]"
+
+
+class HeadlessInteractiveCore:
+    """
+    Headless executor that uses the same interactive core pipeline as TUI.
+    
+    This class reuses:
+    - InteractiveRuntime for ACP/LSP subsystems
+    - get_interactive_tools() for tool resolution
+    - ActionOrchestrator for plan/approve/apply/verify flow
+    
+    Usage:
+        executor = HeadlessInteractiveCore(config)
+        result = executor.run(["Create a file hello.py with print('hello')"])
+        executor.cleanup()  # IMPORTANT: Always call cleanup to stop LSP/ACP
+        
+    Or use as context manager:
+        with HeadlessInteractiveCore(config) as executor:
+            result = executor.run(["Create a file"])
+    """
+    
+    def __init__(self, config: Optional[HeadlessConfig] = None):
+        self.config = config or HeadlessConfig()
+        self._runtime = None
+        self._tools: List[Callable] = []
+        self._tool_trace: List[ToolCallTrace] = []
+        self._agents: List[Any] = []  # Will hold Agent instances
+        self._initialized = False
+        self._cleanup_done = False
+    
+    def __enter__(self):
+        """Context manager entry."""
+        return self
+    
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        """Context manager exit - ensures cleanup."""
+        self.cleanup()
+        return False
+    
+    def __del__(self):
+        """Destructor - attempt cleanup if not done."""
+        if not self._cleanup_done:
+            try:
+                self.cleanup()
+            except Exception:
+                pass  # Best effort cleanup in destructor
+    
+    def _wrap_tool_for_trace(self, tool: Callable) -> Callable:
+        """Wrap a tool to capture calls in trace."""
+        @functools.wraps(tool)
+        def wrapper(*args, **kwargs):
+            start = time.time()
+            timestamp = datetime.utcnow().isoformat()
+            try:
+                result = tool(*args, **kwargs)
+                trace = ToolCallTrace(
+                    tool_name=tool.__name__,
+                    args=args,
+                    kwargs=kwargs,
+                    result=result,
+                    success=True,
+                    error=None,
+                    duration=time.time() - start,
+                    timestamp=timestamp,
+                )
+                self._tool_trace.append(trace)
+                logger.debug(f"Tool {tool.__name__} succeeded in {trace.duration:.3f}s")
+                return result
+            except Exception as e:
+                trace = ToolCallTrace(
+                    tool_name=tool.__name__,
+                    args=args,
+                    kwargs=kwargs,
+                    result=None,
+                    success=False,
+                    error=str(e),
+                    duration=time.time() - start,
+                    timestamp=timestamp,
+                )
+                self._tool_trace.append(trace)
+                logger.debug(f"Tool {tool.__name__} failed: {e}")
+                raise
+        return wrapper
+    
+    def _initialize(self) -> None:
+        """Initialize runtime and tools using interactive core path."""
+        if self._initialized:
+            return
+        
+        logger.debug(f"Initializing headless core with workspace={self.config.workspace}")
+        
+        # Use the SAME tool loading path as interactive TUI
+        # This is from: praisonai/cli/features/interactive_tools.py
+        from praisonai.cli.features.interactive_tools import (
+            get_interactive_tools,
+            ToolConfig,
+        )
+        
+        # Create tool config matching interactive core
+        tool_config = ToolConfig(
+            workspace=self.config.workspace,
+            enable_acp=self.config.enable_acp,
+            enable_lsp=self.config.enable_lsp,
+            enable_basic=self.config.enable_basic,
+            approval_mode=self.config.approval_mode,
+            lsp_enabled=self.config.enable_lsp,
+            acp_enabled=self.config.enable_acp,
+        )
+        
+        # Get tools using the SAME function as TUI
+        raw_tools = get_interactive_tools(config=tool_config)
+        
+        # Wrap tools for tracing
+        self._tools = [self._wrap_tool_for_trace(t) for t in raw_tools]
+        
+        logger.debug(f"Loaded {len(self._tools)} tools: {[t.__name__ for t in self._tools]}")
+        
+        # Create agents using the SAME Agent class as TUI
+        from praisonaiagents import Agent
+        
+        for agent_config in self.config.agents:
+            agent = Agent(
+                name=agent_config.name,
+                role=agent_config.role,
+                goal="Help the user with their requests",
+                instructions=agent_config.instructions,
+                llm=agent_config.llm,
+                tools=self._tools if self._tools else None,
+            )
+            self._agents.append(agent)
+            logger.debug(f"Created agent: {agent_config.name}")
+        
+        self._initialized = True
+    
+    def run(self, prompts: Union[str, List[str]]) -> HeadlessExecutionResult:
+        """
+        Run prompts through headless interactive core.
+        
+        Args:
+            prompts: Single prompt or list of prompts for multi-step execution
+            
+        Returns:
+            HeadlessExecutionResult with responses, tool trace, and transcript
+        """
+        if isinstance(prompts, str):
+            prompts = [prompts]
+        
+        start_time = time.time()
+        self._tool_trace = []  # Reset trace
+        responses = []
+        transcript = []
+        
+        try:
+            self._initialize()
+            
+            # Execute prompts
+            for i, prompt in enumerate(prompts):
+                logger.debug(f"Executing prompt {i+1}/{len(prompts)}: {prompt[:50]}...")
+                transcript.append({"role": "user", "content": prompt})
+                
+                # Determine which agent to use
+                agent_idx = self._get_agent_for_prompt(i, prompt)
+                agent = self._agents[agent_idx]
+                
+                # Execute using the SAME chat method as TUI
+                response = agent.chat(prompt)
+                responses.append(response)
+                transcript.append({"role": "assistant", "content": response})
+                
+                logger.debug(f"Got response: {response[:100] if response else 'None'}...")
+            
+            return HeadlessExecutionResult(
+                success=True,
+                responses=responses,
+                tool_trace=self._tool_trace,
+                duration=time.time() - start_time,
+                transcript=transcript,
+            )
+            
+        except Exception as e:
+            logger.error(f"Headless execution failed: {e}", exc_info=True)
+            return HeadlessExecutionResult(
+                success=False,
+                responses=responses,
+                tool_trace=self._tool_trace,
+                duration=time.time() - start_time,
+                error=str(e),
+                transcript=transcript,
+            )
+    
+    def _get_agent_for_prompt(self, prompt_idx: int, prompt: str) -> int:
+        """Determine which agent should handle a prompt (for multi-agent)."""
+        if len(self._agents) == 1:
+            return 0
+        
+        # If workflow is defined, use it for routing
+        if self.config.workflow:
+            routing = self.config.workflow.get("routing", {})
+            # Simple round-robin or keyword-based routing
+            if "round_robin" in routing:
+                return prompt_idx % len(self._agents)
+            elif "keywords" in routing:
+                for i, agent_config in enumerate(self.config.agents):
+                    keywords = routing["keywords"].get(agent_config.name, [])
+                    if any(kw.lower() in prompt.lower() for kw in keywords):
+                        return i
+        
+        # Default: round-robin
+        return prompt_idx % len(self._agents)
+    
+    def get_tool_trace(self) -> List[Dict[str, Any]]:
+        """Get the tool call trace as list of dicts."""
+        return [t.to_dict() for t in self._tool_trace]
+    
+    def get_tools_called(self) -> List[str]:
+        """Get list of tool names that were called."""
+        return [t.tool_name for t in self._tool_trace]
+    
+    def clear_trace(self) -> None:
+        """Clear the tool trace."""
+        self._tool_trace = []
+    
+    def cleanup(self) -> None:
+        """
+        Clean up resources - MUST be called when done.
+        
+        Stops LSP and ACP subsystems that were started during initialization.
+        This is critical to prevent resource leaks and orphaned processes.
+        """
+        if self._cleanup_done:
+            return
+        
+        logger.debug("Cleaning up HeadlessInteractiveCore resources...")
+        
+        # Stop the runtime if it was created
+        if self._runtime is not None:
+            try:
+                import asyncio
+                
+                # Get or create event loop
+                try:
+                    loop = asyncio.get_running_loop()
+                except RuntimeError:
+                    loop = None
+                
+                if loop and loop.is_running():
+                    # Schedule the stop coroutine
+                    asyncio.ensure_future(self._runtime.stop())
+                else:
+                    # Create new loop and run
+                    asyncio.run(self._runtime.stop())
+                
+                logger.debug("Runtime stopped successfully")
+            except Exception as e:
+                logger.warning(f"Error stopping runtime: {e}")
+            finally:
+                self._runtime = None
+        
+        # Clear agents
+        self._agents = []
+        self._tools = []
+        self._initialized = False
+        self._cleanup_done = True
+        
+        logger.debug("HeadlessInteractiveCore cleanup complete")
+
+
+def run_headless(
+    prompts: Union[str, List[str]],
+    workspace: Optional[str] = None,
+    model: str = "gpt-4o-mini",
+    approval_mode: str = "auto",
+    timeout: int = 60,
+    agents: Optional[List[Dict[str, Any]]] = None,
+    workflow: Optional[Dict[str, Any]] = None,
+) -> HeadlessExecutionResult:
+    """
+    Convenience function to run prompts through headless interactive core.
+    
+    Args:
+        prompts: Single prompt or list of prompts
+        workspace: Working directory (default: cwd)
+        model: LLM model to use
+        approval_mode: Approval mode (auto, manual, scoped)
+        timeout: Timeout in seconds
+        agents: List of agent configs for multi-agent
+        workflow: Workflow config for turn routing
+        
+    Returns:
+        HeadlessExecutionResult
+        
+    Example:
+        result = run_headless("Create a file hello.py with print('hello')")
+        print(result.responses[0])
+        print(result.get_tools_called())
+    """
+    agent_configs = []
+    if agents:
+        for a in agents:
+            agent_configs.append(AgentConfig(
+                name=a.get("name", "Agent"),
+                instructions=a.get("instructions", "You are a helpful assistant."),
+                role=a.get("role", "assistant"),
+                llm=a.get("llm", model),
+            ))
+    
+    config = HeadlessConfig(
+        workspace=workspace or os.getcwd(),
+        model=model,
+        approval_mode=approval_mode,
+        timeout=timeout,
+        agents=agent_configs if agent_configs else None,
+        workflow=workflow,
+    )
+    
+    # Use context manager to ensure cleanup
+    with HeadlessInteractiveCore(config) as executor:
+        return executor.run(prompts)

src/praisonai/praisonai/cli/features/interactive_runtime.py

@@ -40,7 +40,7 @@ class RuntimeConfig:
     workspace: str = "."
     lsp_enabled: bool = True
     acp_enabled: bool = True
-    approval_mode: str = "manual"  # manual, auto, scoped
+    approval_mode: str = "auto"  # auto (full privileges), manual, scoped
     trace_enabled: bool = False
     trace_file: Optional[str] = None
     json_output: bool = False
@@ -146,15 +146,21 @@ class InteractiveRuntime:
     def read_only(self) -> bool:
         """Check if runtime is in read-only mode.
         
-        Note: When PRAISON_APPROVAL_MODE=auto, we allow writes even without ACP
-        to support non-interactive automation and testing scenarios.
+        Note: When approval_mode=auto (via config or env), we allow writes even 
+        without ACP to support non-interactive automation and testing scenarios.
         """
-        # Auto-approval mode bypasses read-only for automation
-        import os as _os
-        approval_mode = _os.environ.get("PRAISON_APPROVAL_MODE", "").lower()
-        if approval_mode == "auto":
+        # Check config approval_mode first (takes precedence)
+        if self.config.approval_mode == "auto":
             return False
-        return self._read_only or not self.acp_ready
+        
+        # Then check environment variable as fallback
+        import os as _os
+        env_approval_mode = _os.environ.get("PRAISON_APPROVAL_MODE", "").lower()
+        if env_approval_mode == "auto":
+            return False
+        
+        # Only enforce read-only if explicitly set or ACP not ready
+        return self._read_only
     
     def get_status(self) -> Dict[str, Any]:
         """Get runtime status."""

src/praisonai/praisonai/cli/features/interactive_test_harness.py

@@ -0,0 +1,537 @@
+"""
+Interactive Test Harness for PraisonAI.
+
+Provides workspace isolation, fixtures, tool trace wrappers, and artifact writing
+for testing interactive mode through the headless core executor.
+"""
+
+import json
+import logging
+import re
+import shutil
+import tempfile
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+logger = logging.getLogger(__name__)
+
+
+# Blocked commands for safe execution
+BLOCKED_COMMANDS = {
+    "rm -rf /",
+    "rm -rf /*",
+    "sudo rm",
+    "mkfs",
+    "dd if=",
+    ":(){:|:&};:",  # Fork bomb
+    "chmod -R 777 /",
+    "curl | sh",
+    "wget | sh",
+}
+
+# Allowed commands (whitelist for strict mode)
+SAFE_COMMANDS = {
+    "echo",
+    "cat",
+    "ls",
+    "pwd",
+    "head",
+    "tail",
+    "wc",
+    "grep",
+    "find",
+    "python",
+    "python3",
+    "pip",
+    "pip3",
+    "node",
+    "npm",
+}
+
+
+@dataclass
+class WorkspaceFixture:
+    """Definition of a workspace fixture."""
+    name: str
+    files: Dict[str, str] = field(default_factory=dict)  # path -> content
+    directories: List[str] = field(default_factory=list)
+    
+    @classmethod
+    def empty(cls) -> "WorkspaceFixture":
+        """Create an empty workspace fixture."""
+        return cls(name="empty")
+    
+    @classmethod
+    def seeded(cls) -> "WorkspaceFixture":
+        """Create a seeded workspace with common files."""
+        return cls(
+            name="seeded",
+            files={
+                "README.md": "# Test Project\n\nThis is a test project.",
+                "main.py": "# Main entry point\nprint('Hello, World!')\n",
+                "config.json": '{"name": "test", "version": "1.0.0"}',
+            },
+            directories=["src", "tests"],
+        )
+    
+    @classmethod
+    def git(cls) -> "WorkspaceFixture":
+        """Create a git-initialized workspace."""
+        fixture = cls.seeded()
+        fixture.name = "git"
+        fixture.files[".gitignore"] = "__pycache__/\n*.pyc\n.env\n"
+        return fixture
+    
+    @classmethod
+    def python_project(cls) -> "WorkspaceFixture":
+        """Create a Python project workspace."""
+        return cls(
+            name="python_project",
+            files={
+                "README.md": "# Python Project\n",
+                "main.py": '''"""Main module."""
+
+def main():
+    """Main function."""
+    print("Hello from main!")
+
+if __name__ == "__main__":
+    main()
+''',
+                "utils.py": '''"""Utility functions."""
+
+def add(a: int, b: int) -> int:
+    """Add two numbers."""
+    return a + b
+
+def multiply(a: int, b: int) -> int:
+    """Multiply two numbers."""
+    return a * b
+
+class Calculator:
+    """Simple calculator class."""
+    
+    def __init__(self):
+        self.result = 0
+    
+    def add(self, x: int) -> int:
+        self.result += x
+        return self.result
+''',
+                "requirements.txt": "# No dependencies\n",
+            },
+            directories=["tests"],
+        )
+
+
+BUILTIN_FIXTURES = {
+    "empty": WorkspaceFixture.empty,
+    "seeded": WorkspaceFixture.seeded,
+    "git": WorkspaceFixture.git,
+    "python_project": WorkspaceFixture.python_project,
+}
+
+
+@dataclass
+class TestArtifacts:
+    """Container for test artifacts."""
+    transcript: List[Dict[str, str]] = field(default_factory=list)
+    tool_trace: List[Dict[str, Any]] = field(default_factory=list)
+    result: Dict[str, Any] = field(default_factory=dict)
+    workspace_snapshot: Optional[Dict[str, str]] = None
+    judge_result: Optional[Dict[str, Any]] = None
+    
+    def save(self, artifacts_dir: Path) -> None:
+        """Save artifacts to directory."""
+        artifacts_dir.mkdir(parents=True, exist_ok=True)
+        
+        # Save transcript
+        transcript_path = artifacts_dir / "transcript.txt"
+        with open(transcript_path, "w") as f:
+            for entry in self.transcript:
+                role = entry.get("role", "unknown")
+                content = entry.get("content", "")
+                f.write(f"[{role.upper()}]\n{content}\n\n")
+        
+        # Save tool trace as JSONL
+        trace_path = artifacts_dir / "tool_trace.jsonl"
+        with open(trace_path, "w") as f:
+            for trace in self.tool_trace:
+                f.write(json.dumps(trace) + "\n")
+        
+        # Save result
+        result_path = artifacts_dir / "result.json"
+        with open(result_path, "w") as f:
+            json.dump(self.result, f, indent=2, default=str)
+        
+        # Save workspace snapshot if present
+        if self.workspace_snapshot:
+            snapshot_dir = artifacts_dir / "workspace"
+            snapshot_dir.mkdir(exist_ok=True)
+            for rel_path, content in self.workspace_snapshot.items():
+                file_path = snapshot_dir / rel_path
+                file_path.parent.mkdir(parents=True, exist_ok=True)
+                file_path.write_text(content)
+        
+        # Save judge result if present
+        if self.judge_result:
+            judge_path = artifacts_dir / "judge_result.json"
+            with open(judge_path, "w") as f:
+                json.dump(self.judge_result, f, indent=2, default=str)
+        
+        logger.debug(f"Saved artifacts to {artifacts_dir}")
+
+
+class InteractiveTestHarness:
+    """
+    Test harness for interactive mode testing.
+    
+    Provides:
+    - Workspace isolation (temp directory)
+    - Fixture setup
+    - Tool trace capture
+    - Artifact generation
+    - File assertions
+    - Tool call assertions
+    """
+    
+    def __init__(
+        self,
+        workspace: Optional[Path] = None,
+        artifacts_dir: Optional[Path] = None,
+        keep_workspace: bool = False,
+        safe_commands: bool = True,
+    ):
+        """
+        Initialize test harness.
+        
+        Args:
+            workspace: Workspace directory (created if None)
+            artifacts_dir: Directory for artifacts (created if None)
+            keep_workspace: Keep workspace after cleanup
+            safe_commands: Enforce safe command execution
+        """
+        self._workspace_created = workspace is None
+        self.workspace = workspace or Path(tempfile.mkdtemp(prefix="praison_test_"))
+        self.artifacts_dir = artifacts_dir or self.workspace / "artifacts"
+        self.keep_workspace = keep_workspace
+        self.safe_commands = safe_commands
+        self.artifacts = TestArtifacts()
+        self._executor = None
+    
+    def setup_workspace(self, fixture: str = "empty") -> None:
+        """
+        Set up workspace with fixture.
+        
+        Args:
+            fixture: Fixture name (empty, seeded, git, python_project)
+        """
+        # Get fixture
+        if fixture in BUILTIN_FIXTURES:
+            fixture_obj = BUILTIN_FIXTURES[fixture]()
+        else:
+            logger.warning(f"Unknown fixture '{fixture}', using empty")
+            fixture_obj = WorkspaceFixture.empty()
+        
+        # Create directories
+        self.workspace.mkdir(parents=True, exist_ok=True)
+        for dir_path in fixture_obj.directories:
+            (self.workspace / dir_path).mkdir(parents=True, exist_ok=True)
+        
+        # Create files
+        for rel_path, content in fixture_obj.files.items():
+            file_path = self.workspace / rel_path
+            file_path.parent.mkdir(parents=True, exist_ok=True)
+            file_path.write_text(content)
+        
+        # Initialize git if needed
+        if fixture == "git":
+            import subprocess
+            try:
+                subprocess.run(
+                    ["git", "init"],
+                    cwd=self.workspace,
+                    capture_output=True,
+                    timeout=10,
+                )
+            except Exception as e:
+                logger.debug(f"Git init failed: {e}")
+        
+        logger.debug(f"Set up workspace with fixture '{fixture}' at {self.workspace}")
+    
+    def create_executor(
+        self,
+        model: str = "gpt-4o-mini",
+        approval_mode: str = "auto",
+        agents: Optional[List[Dict[str, Any]]] = None,
+        workflow: Optional[Dict[str, Any]] = None,
+    ):
+        """
+        Create headless executor using interactive core.
+        
+        Args:
+            model: LLM model
+            approval_mode: Approval mode (auto, manual, scoped)
+            agents: Multi-agent configs
+            workflow: Workflow routing config
+            
+        Returns:
+            HeadlessInteractiveCore executor
+        """
+        from .interactive_core_headless import (
+            HeadlessConfig,
+            HeadlessInteractiveCore,
+            AgentConfig,
+        )
+        
+        agent_configs = []
+        if agents:
+            for a in agents:
+                agent_configs.append(AgentConfig(
+                    name=a.get("name", "Agent"),
+                    instructions=a.get("instructions", "You are a helpful assistant."),
+                    role=a.get("role", "assistant"),
+                    llm=a.get("llm", model),
+                ))
+        
+        config = HeadlessConfig(
+            workspace=str(self.workspace),
+            model=model,
+            approval_mode=approval_mode,
+            agents=agent_configs if agent_configs else None,
+            workflow=workflow,
+        )
+        
+        self._executor = HeadlessInteractiveCore(config)
+        return self._executor
+    
+    def run(
+        self,
+        prompts: List[str],
+        model: str = "gpt-4o-mini",
+        approval_mode: str = "auto",
+        agents: Optional[List[Dict[str, Any]]] = None,
+        workflow: Optional[Dict[str, Any]] = None,
+    ):
+        """
+        Run prompts through headless interactive core.
+        
+        Args:
+            prompts: List of prompts to execute
+            model: LLM model
+            approval_mode: Approval mode
+            agents: Multi-agent configs
+            workflow: Workflow routing
+            
+        Returns:
+            HeadlessExecutionResult
+        """
+        if not self._executor:
+            self.create_executor(model, approval_mode, agents, workflow)
+        
+        result = self._executor.run(prompts)
+        
+        # Store in artifacts
+        self.artifacts.transcript = result.transcript
+        self.artifacts.tool_trace = self._executor.get_tool_trace()
+        self.artifacts.result = result.to_dict()
+        
+        return result
+    
+    def verify_files(self, expected_files: Dict[str, str]) -> Dict[str, bool]:
+        """
+        Verify expected files exist and match content patterns.
+        
+        Args:
+            expected_files: Dict of {relative_path: content_regex}
+            
+        Returns:
+            Dict of {path: passed}
+        """
+        results = {}
+        
+        for rel_path, content_pattern in expected_files.items():
+            file_path = self.workspace / rel_path
+            
+            if not file_path.exists():
+                logger.debug(f"File not found: {rel_path}")
+                results[rel_path] = False
+                continue
+            
+            content = file_path.read_text()
+            
+            if content_pattern:
+                try:
+                    if re.search(content_pattern, content, re.MULTILINE | re.DOTALL):
+                        results[rel_path] = True
+                    else:
+                        logger.debug(f"Content mismatch for {rel_path}: pattern '{content_pattern}' not found")
+                        results[rel_path] = False
+                except re.error as e:
+                    logger.warning(f"Invalid regex pattern for {rel_path}: {e}")
+                    results[rel_path] = False
+            else:
+                # Just check existence
+                results[rel_path] = True
+        
+        return results
+    
+    def verify_tool_calls(
+        self,
+        expected_tools: Optional[List[str]] = None,
+        forbidden_tools: Optional[List[str]] = None,
+    ) -> Dict[str, Any]:
+        """
+        Verify tool calls match expectations.
+        
+        Args:
+            expected_tools: Tools that MUST have been called
+            forbidden_tools: Tools that MUST NOT have been called
+            
+        Returns:
+            Dict with verification results
+        """
+        called_tools = set()
+        if self._executor:
+            called_tools = set(self._executor.get_tools_called())
+        
+        results = {
+            "called_tools": list(called_tools),
+            "expected_passed": True,
+            "forbidden_passed": True,
+            "missing_tools": [],
+            "forbidden_called": [],
+        }
+        
+        # Check expected tools
+        if expected_tools:
+            expected_set = set(expected_tools)
+            missing = expected_set - called_tools
+            if missing:
+                results["expected_passed"] = False
+                results["missing_tools"] = list(missing)
+        
+        # Check forbidden tools
+        if forbidden_tools:
+            forbidden_set = set(forbidden_tools)
+            forbidden_called = forbidden_set & called_tools
+            if forbidden_called:
+                results["forbidden_passed"] = False
+                results["forbidden_called"] = list(forbidden_called)
+        
+        results["passed"] = results["expected_passed"] and results["forbidden_passed"]
+        return results
+    
+    def verify_response(self, response: str, pattern: str) -> bool:
+        """
+        Verify response matches pattern.
+        
+        Args:
+            response: Response text
+            pattern: Regex pattern
+            
+        Returns:
+            True if matches
+        """
+        if not pattern:
+            return True
+        
+        try:
+            return bool(re.search(pattern, response, re.MULTILINE | re.DOTALL | re.IGNORECASE))
+        except re.error as e:
+            logger.warning(f"Invalid response pattern: {e}")
+            return False
+    
+    def snapshot_workspace(self) -> Dict[str, str]:
+        """
+        Create snapshot of workspace files.
+        
+        Returns:
+            Dict of {relative_path: content}
+        """
+        snapshot = {}
+        
+        for file_path in self.workspace.rglob("*"):
+            if file_path.is_file():
+                rel_path = str(file_path.relative_to(self.workspace))
+                # Skip binary files and large files
+                if file_path.suffix in {".pyc", ".so", ".dll", ".exe", ".bin"}:
+                    continue
+                if file_path.stat().st_size > 100_000:  # 100KB limit
+                    continue
+                try:
+                    snapshot[rel_path] = file_path.read_text()
+                except Exception:
+                    pass  # Skip unreadable files
+        
+        self.artifacts.workspace_snapshot = snapshot
+        return snapshot
+    
+    def save_artifacts(self, test_id: str = "test") -> Path:
+        """
+        Save all artifacts to directory.
+        
+        Args:
+            test_id: Test identifier for subdirectory
+            
+        Returns:
+            Path to artifacts directory
+        """
+        test_artifacts_dir = self.artifacts_dir / test_id
+        self.artifacts.save(test_artifacts_dir)
+        return test_artifacts_dir
+    
+    def cleanup(self) -> None:
+        """Clean up workspace and executor resources."""
+        # Clean up executor (stops LSP/ACP subsystems)
+        if self._executor is not None:
+            try:
+                self._executor.cleanup()
+                logger.debug("Cleaned up executor resources")
+            except Exception as e:
+                logger.warning(f"Failed to clean up executor: {e}")
+            finally:
+                self._executor = None
+        
+        # Clean up workspace
+        if self._workspace_created and not self.keep_workspace:
+            try:
+                shutil.rmtree(self.workspace)
+                logger.debug(f"Cleaned up workspace: {self.workspace}")
+            except Exception as e:
+                logger.warning(f"Failed to clean up workspace: {e}")
+    
+    def __enter__(self):
+        return self
+    
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        self.cleanup()
+        return False
+
+
+def create_test_harness(
+    fixture: str = "empty",
+    workspace: Optional[Path] = None,
+    artifacts_dir: Optional[Path] = None,
+    keep_workspace: bool = False,
+) -> InteractiveTestHarness:
+    """
+    Create and set up a test harness.
+    
+    Args:
+        fixture: Workspace fixture name
+        workspace: Workspace path (temp if None)
+        artifacts_dir: Artifacts directory
+        keep_workspace: Keep workspace after cleanup
+        
+    Returns:
+        Configured InteractiveTestHarness
+    """
+    harness = InteractiveTestHarness(
+        workspace=workspace,
+        artifacts_dir=artifacts_dir,
+        keep_workspace=keep_workspace,
+    )
+    harness.setup_workspace(fixture)
+    return harness

src/praisonai/praisonai/cli/features/interactive_tools.py

@@ -72,7 +72,7 @@ class ToolConfig:
     enable_acp: bool = True
     enable_lsp: bool = True
     enable_basic: bool = True
-    approval_mode: str = "auto"  # auto, manual, scoped
+    approval_mode: str = "auto"  # auto (full privileges), manual, scoped
     lsp_enabled: bool = True
     acp_enabled: bool = True
     

src/praisonai/praisonai/cli/interactive/config.py

@@ -50,8 +50,9 @@ class InteractiveConfig:
     autonomy: bool = True  # Enable autonomy by default for complex tasks
     autonomy_config: Optional[dict] = None  # Custom autonomy config
     
-    # Approval mode
-    approval_mode: Union[str, ApprovalMode] = "prompt"
+    # Approval mode - default to "auto" for full privileges in interactive mode
+    # Use "prompt" for safer mode that asks before each action
+    approval_mode: Union[str, ApprovalMode] = "auto"
     
     # File attachments
     files: List[str] = field(default_factory=list)

src/praisonai/praisonai/cli/interactive/events.py

@@ -24,6 +24,9 @@ class InteractiveEventType(Enum):
     TOOL_START = "tool.start"
     TOOL_END = "tool.end"
     
+    # Stage transitions (PLAN → EXPLORE → BUILD → REVIEW)
+    STAGE_CHANGE = "stage.change"
+    
     # Approval/permission flow
     APPROVAL_ASKED = "approval.asked"
     APPROVAL_ANSWERED = "approval.answered"

src/praisonai/praisonai/cli/interactive/praison_io.py

@@ -0,0 +1,502 @@
+"""
+PraisonIO - Interactive CLI I/O handler inspired by Aider.
+
+Key design principles:
+1. prompt_toolkit used SYNCHRONOUSLY (never inside asyncio.run)
+2. Rich for output rendering
+3. MarkdownStream for flicker-free streaming
+4. Clean separation of input/output concerns
+"""
+
+import sys
+import time
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Dict, List, Optional, Set
+
+# Lazy imports for performance
+_PROMPT_TOOLKIT_AVAILABLE = None
+_RICH_AVAILABLE = None
+
+
+def _check_prompt_toolkit():
+    global _PROMPT_TOOLKIT_AVAILABLE
+    if _PROMPT_TOOLKIT_AVAILABLE is None:
+        try:
+            import prompt_toolkit  # noqa: F401
+            _PROMPT_TOOLKIT_AVAILABLE = True
+        except ImportError:
+            _PROMPT_TOOLKIT_AVAILABLE = False
+    return _PROMPT_TOOLKIT_AVAILABLE
+
+
+def _check_rich():
+    global _RICH_AVAILABLE
+    if _RICH_AVAILABLE is None:
+        try:
+            import rich  # noqa: F401
+            _RICH_AVAILABLE = True
+        except ImportError:
+            _RICH_AVAILABLE = False
+    return _RICH_AVAILABLE
+
+
+@dataclass
+class IOConfig:
+    """Configuration for PraisonIO."""
+    # Appearance
+    pretty: bool = True
+    user_input_color: str = "#00cc00"
+    tool_output_color: str = "#0088ff"
+    tool_error_color: str = "#ff0000"
+    tool_warning_color: str = "#ffaa00"
+    assistant_color: str = "#00aaff"
+    
+    # Input settings
+    multiline_mode: bool = False
+    vi_mode: bool = False
+    
+    # History
+    input_history_file: Optional[str] = None
+    
+    # Completion
+    enable_completions: bool = True
+    
+    # Streaming
+    stream_delay: float = 0.05  # 20fps default
+
+
+class PraisonCompleter:
+    """Auto-completer for commands, files, and symbols."""
+    
+    def __init__(self, commands: Optional[Dict[str, str]] = None):
+        self.commands = commands or {}
+        self.files: Set[str] = set()
+        self.symbols: Set[str] = set()
+    
+    def add_files(self, files: List[str]) -> None:
+        """Add files to completion."""
+        self.files.update(files)
+    
+    def add_symbols(self, symbols: List[str]) -> None:
+        """Add symbols to completion."""
+        self.symbols.update(symbols)
+    
+    def get_completions(self, text: str, cursor_pos: int) -> List[tuple]:
+        """Get completions for current input."""
+        before_cursor = text[:cursor_pos]
+        words = before_cursor.split()
+        current_word = words[-1] if words else ""
+        
+        completions = []
+        
+        # Slash commands
+        if current_word.startswith("/"):
+            cmd_part = current_word[1:].lower()
+            for cmd, desc in self.commands.items():
+                if cmd.lower().startswith(cmd_part):
+                    completions.append((f"/{cmd}", desc))
+        
+        # @ file mentions
+        elif current_word.startswith("@"):
+            file_part = current_word[1:].lower()
+            for f in self.files:
+                if file_part in f.lower():
+                    completions.append((f"@{f}", "file"))
+        
+        # General word completion
+        elif current_word:
+            word_lower = current_word.lower()
+            for symbol in self.symbols:
+                if symbol.lower().startswith(word_lower):
+                    completions.append((symbol, "symbol"))
+        
+        return completions[:20]
+    
+    def create_prompt_toolkit_completer(self):
+        """Create prompt_toolkit Completer."""
+        if not _check_prompt_toolkit():
+            return None
+        
+        from prompt_toolkit.completion import Completer, Completion
+        
+        outer = self
+        
+        class PTCompleter(Completer):
+            def get_completions(self, document, complete_event):
+                text = document.text
+                cursor_pos = document.cursor_position
+                
+                for completion, meta in outer.get_completions(text, cursor_pos):
+                    words = text[:cursor_pos].split()
+                    current_word = words[-1] if words else ""
+                    start_pos = -len(current_word)
+                    yield Completion(completion, start_position=start_pos, display_meta=meta)
+        
+        return PTCompleter()
+
+
+class MarkdownStream:
+    """Flicker-free streaming markdown renderer.
+    
+    Uses Rich Live with a sliding window approach:
+    - Stable lines print above the live area
+    - Unstable lines (last N) render in the live window
+    """
+    
+    def __init__(self, live_window: int = 6):
+        self.live_window = live_window
+        self.printed: List[str] = []
+        self.live = None
+        self._live_started = False
+        self.min_delay = 1.0 / 20  # 20fps
+        self.when = 0
+    
+    def _render_to_lines(self, text: str) -> List[str]:
+        """Render markdown to lines."""
+        if not _check_rich():
+            return text.splitlines(keepends=True)
+        
+        import io
+        from rich.console import Console
+        from rich.markdown import Markdown
+        
+        string_io = io.StringIO()
+        console = Console(file=string_io, force_terminal=True, width=100)
+        console.print(Markdown(text))
+        return string_io.getvalue().splitlines(keepends=True)
+    
+    def update(self, text: str, final: bool = False) -> None:
+        """Update the streaming display."""
+        if not _check_rich():
+            # Fallback: just print
+            print(text, end="" if not final else "\n", flush=True)
+            return
+        
+        from rich.live import Live
+        from rich.text import Text
+        
+        # Start live on first update
+        if not self._live_started:
+            self.live = Live(Text(""), refresh_per_second=20)
+            self.live.start()
+            self._live_started = True
+        
+        # Throttle updates
+        now = time.time()
+        if not final and now - self.when < self.min_delay:
+            return
+        self.when = now
+        
+        lines = self._render_to_lines(text)
+        num_lines = len(lines)
+        
+        # Calculate stable lines
+        if not final:
+            num_lines -= self.live_window
+        
+        if final or num_lines > 0:
+            num_printed = len(self.printed)
+            show_count = num_lines - num_printed
+            
+            if show_count > 0:
+                show = lines[num_printed:num_lines]
+                show_text = "".join(show)
+                self.live.console.print(Text.from_ansi(show_text), end="")
+                self.printed = lines[:num_lines]
+        
+        if final:
+            self.live.update(Text(""))
+            self.live.stop()
+            self.live = None
+            return
+        
+        # Update live window
+        rest = lines[num_lines:]
+        rest_text = "".join(rest)
+        self.live.update(Text.from_ansi(rest_text))
+    
+    def __del__(self):
+        if self.live:
+            try:
+                self.live.stop()
+            except Exception:
+                pass
+
+
+class PraisonIO:
+    """
+    Main I/O handler for PraisonAI interactive CLI.
+    
+    Inspired by Aider's io.py - handles all terminal I/O with:
+    - Rich formatting for output
+    - prompt_toolkit for input (with completions, history)
+    - Streaming markdown rendering
+    """
+    
+    def __init__(self, config: Optional[IOConfig] = None):
+        self.config = config or IOConfig()
+        self.completer = PraisonCompleter()
+        self._console = None
+        self._prompt_session = None
+        self._history_file = None
+        
+        # Setup history file
+        if self.config.input_history_file:
+            self._history_file = Path(self.config.input_history_file)
+        else:
+            self._history_file = Path.home() / ".praison" / "input_history"
+        
+        self._history_file.parent.mkdir(parents=True, exist_ok=True)
+    
+    @property
+    def console(self):
+        """Lazy-load Rich console."""
+        if self._console is None and _check_rich():
+            from rich.console import Console
+            self._console = Console()
+        return self._console
+    
+    def _create_prompt_session(self):
+        """Create prompt_toolkit session."""
+        if not _check_prompt_toolkit():
+            return None
+        
+        from prompt_toolkit import PromptSession
+        from prompt_toolkit.history import FileHistory
+        from prompt_toolkit.auto_suggest import AutoSuggestFromHistory
+        from prompt_toolkit.key_binding import KeyBindings
+        
+        kb = KeyBindings()
+        
+        # Ctrl+C to cancel current input
+        @kb.add("c-c")
+        def _(event):
+            event.current_buffer.reset()
+        
+        # Alt+Enter for newline in multiline mode
+        @kb.add("escape", "enter")
+        def _(event):
+            event.current_buffer.insert_text("\n")
+        
+        # Enter to submit (unless in multiline mode)
+        @kb.add("enter")
+        def _(event):
+            if self.config.multiline_mode:
+                event.current_buffer.insert_text("\n")
+            else:
+                event.current_buffer.validate_and_handle()
+        
+        # Ctrl+D to submit in multiline mode
+        @kb.add("c-d")
+        def _(event):
+            event.current_buffer.validate_and_handle()
+        
+        session_kwargs = {
+            "history": FileHistory(str(self._history_file)),
+            "auto_suggest": AutoSuggestFromHistory(),
+            "key_bindings": kb,
+            "multiline": self.config.multiline_mode,
+        }
+        
+        if self.config.enable_completions:
+            completer = self.completer.create_prompt_toolkit_completer()
+            if completer:
+                from prompt_toolkit.completion import ThreadedCompleter
+                session_kwargs["completer"] = ThreadedCompleter(completer)
+        
+        if self.config.vi_mode:
+            from prompt_toolkit.enums import EditingMode
+            session_kwargs["editing_mode"] = EditingMode.VI
+        
+        return PromptSession(**session_kwargs)
+    
+    def _get_style(self):
+        """Get prompt_toolkit style."""
+        if not _check_prompt_toolkit() or not self.config.pretty:
+            return None
+        
+        from prompt_toolkit.styles import Style
+        
+        style_dict = {}
+        if self.config.user_input_color:
+            style_dict[""] = self.config.user_input_color
+        
+        return Style.from_dict(style_dict) if style_dict else None
+    
+    def get_input(self, prompt: str = "You: ") -> str:
+        """
+        Get input from user.
+        
+        This is SYNCHRONOUS - never call from inside asyncio.run().
+        """
+        if self._prompt_session is None:
+            self._prompt_session = self._create_prompt_session()
+        
+        if self._prompt_session:
+            try:
+                style = self._get_style()
+                return self._prompt_session.prompt(prompt, style=style)
+            except KeyboardInterrupt:
+                return ""
+            except EOFError:
+                return "/exit"
+        else:
+            # Fallback to basic input
+            try:
+                return input(prompt)
+            except KeyboardInterrupt:
+                return ""
+            except EOFError:
+                return "/exit"
+    
+    def rule(self) -> None:
+        """Print a horizontal rule."""
+        if self.console and self.config.pretty:
+            self.console.rule(style=self.config.user_input_color)
+        else:
+            print()
+    
+    def print_welcome(self, model: str = "gpt-4o-mini", tools_count: int = 0, session_id: str = "") -> None:
+        """Print welcome message."""
+        if self.console and self.config.pretty:
+            from rich.panel import Panel
+            from rich.text import Text
+            
+            welcome = Text()
+            welcome.append("PraisonAI Interactive Mode\n", style="bold cyan")
+            welcome.append("Type your prompt, use ", style="dim")
+            welcome.append("/help", style="bold green")
+            welcome.append(" for commands, ", style="dim")
+            welcome.append("/exit", style="bold yellow")
+            welcome.append(" to quit\n", style="dim")
+            welcome.append(f"Model: {model}", style="dim cyan")
+            if tools_count:
+                welcome.append(f" | Tools: {tools_count}", style="dim")
+            if session_id:
+                welcome.append(f" | Session: {session_id[:8]}", style="dim")
+            
+            self.console.print(Panel(welcome, border_style="blue", padding=(0, 1)))
+        else:
+            print("PraisonAI Interactive Mode")
+            print("Type your prompt, use /help for commands, /exit to quit")
+            print(f"Model: {model} | Tools: {tools_count}")
+    
+    def print_user_input(self, text: str) -> None:
+        """Echo user input with styling."""
+        if self.console and self.config.pretty:
+            from rich.text import Text
+            self.console.print(Text(text, style=self.config.user_input_color))
+    
+    def print_assistant_start(self) -> None:
+        """Print assistant response start marker."""
+        if self.console and self.config.pretty:
+            self.console.print("\n[bold cyan]Assistant:[/bold cyan]")
+        else:
+            print("\nAssistant:")
+    
+    def print_assistant_response(self, response: str) -> None:
+        """Print assistant response with markdown formatting."""
+        if self.console and self.config.pretty:
+            from rich.markdown import Markdown
+            from rich.panel import Panel
+            
+            try:
+                md = Markdown(response)
+                self.console.print(Panel(md, border_style="cyan", padding=(0, 1)))
+            except Exception:
+                self.console.print(response)
+        else:
+            print(response)
+    
+    def stream_response(self, text: str, final: bool = False) -> MarkdownStream:
+        """Get or update streaming response."""
+        if not hasattr(self, '_current_stream') or self._current_stream is None:
+            self._current_stream = MarkdownStream()
+        
+        self._current_stream.update(text, final=final)
+        
+        if final:
+            self._current_stream = None
+        
+        return self._current_stream
+    
+    def tool_output(self, message: str) -> None:
+        """Print tool output."""
+        if self.console and self.config.pretty:
+            self.console.print(f"[{self.config.tool_output_color}]⚙ {message}[/]")
+        else:
+            print(f"⚙ {message}")
+    
+    def tool_error(self, message: str) -> None:
+        """Print tool error."""
+        if self.console and self.config.pretty:
+            self.console.print(f"[{self.config.tool_error_color}]✗ {message}[/]")
+        else:
+            print(f"✗ {message}", file=sys.stderr)
+    
+    def tool_warning(self, message: str) -> None:
+        """Print tool warning."""
+        if self.console and self.config.pretty:
+            self.console.print(f"[{self.config.tool_warning_color}]⚠ {message}[/]")
+        else:
+            print(f"⚠ {message}", file=sys.stderr)
+    
+    def info(self, message: str) -> None:
+        """Print info message."""
+        if self.console and self.config.pretty:
+            self.console.print(f"[dim]ℹ {message}[/dim]")
+        else:
+            print(f"ℹ {message}")
+    
+    def success(self, message: str) -> None:
+        """Print success message."""
+        if self.console and self.config.pretty:
+            self.console.print(f"[green]✓ {message}[/green]")
+        else:
+            print(f"✓ {message}")
+    
+    def print_help(self, commands: Dict[str, str]) -> None:
+        """Print help for available commands."""
+        if self.console and self.config.pretty:
+            from rich.table import Table
+            
+            table = Table(title="Available Commands", border_style="blue")
+            table.add_column("Command", style="green")
+            table.add_column("Description", style="dim")
+            
+            for cmd, desc in sorted(commands.items()):
+                table.add_row(f"/{cmd}", desc)
+            
+            self.console.print(table)
+        else:
+            print("\nAvailable Commands:")
+            for cmd, desc in sorted(commands.items()):
+                print(f"  /{cmd:<12} - {desc}")
+    
+    def confirm(self, message: str, default: bool = False) -> bool:
+        """Ask for confirmation."""
+        suffix = " [Y/n]: " if default else " [y/N]: "
+        
+        if self.console and self.config.pretty:
+            self.console.print(f"[yellow]? {message}[/yellow]", end="")
+        
+        try:
+            response = input(suffix).strip().lower()
+            if not response:
+                return default
+            return response in ("y", "yes")
+        except (EOFError, KeyboardInterrupt):
+            return False
+    
+    def add_commands(self, commands: Dict[str, str]) -> None:
+        """Add commands for completion."""
+        self.completer.commands.update(commands)
+    
+    def add_files(self, files: List[str]) -> None:
+        """Add files for completion."""
+        self.completer.add_files(files)
+    
+    def add_symbols(self, symbols: List[str]) -> None:
+        """Add symbols for completion."""
+        self.completer.add_symbols(symbols)

src/praisonai/praisonai/cli/interactive/repl.py

@@ -0,0 +1,297 @@
+"""
+Interactive REPL for PraisonAI.
+
+This is the main entry point for interactive chat mode.
+Uses PraisonIO for input/output (prompt_toolkit is used SYNCHRONOUSLY).
+"""
+
+import logging
+from dataclasses import dataclass
+from typing import Dict, List, Optional
+
+from .praison_io import PraisonIO, IOConfig
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class REPLConfig:
+    """Configuration for the interactive REPL."""
+    model: str = "gpt-4o-mini"
+    verbose: bool = False
+    memory: bool = False
+    tools: Optional[List] = None
+    session_id: Optional[str] = None
+    continue_session: bool = False
+    workspace: Optional[str] = None
+    autonomy: bool = True
+
+
+# Default slash commands
+DEFAULT_COMMANDS: Dict[str, str] = {
+    "help": "Show available commands",
+    "exit": "Exit interactive mode",
+    "quit": "Exit interactive mode",
+    "clear": "Clear conversation history",
+    "session": "Show current session info",
+    "model": "Show or change model",
+    "cost": "Show token usage and cost",
+    "history": "Show conversation history",
+    "export": "Export conversation to file",
+    "compact": "Toggle compact output mode",
+    "multiline": "Toggle multiline input mode",
+}
+
+
+class InteractiveREPL:
+    """
+    Interactive REPL for PraisonAI chat.
+    
+    Key design:
+    - Uses PraisonIO for all I/O (prompt_toolkit used synchronously)
+    - Agent execution is async but input is sync
+    - Clean separation between I/O and agent logic
+    """
+    
+    def __init__(
+        self,
+        config: Optional[REPLConfig] = None,
+        io_config: Optional[IOConfig] = None,
+    ):
+        self.config = config or REPLConfig()
+        self.io = PraisonIO(io_config)
+        self._agent = None
+        self._running = False
+        self._conversation_history: List[Dict] = []
+        self._session_id: Optional[str] = None
+        self._total_tokens = 0
+        self._total_cost = 0.0
+        
+        # Setup commands
+        self.io.add_commands(DEFAULT_COMMANDS)
+    
+    def _get_agent(self):
+        """Lazy-load the agent."""
+        if self._agent is None:
+            try:
+                from praisonaiagents import Agent
+                
+                agent_kwargs = {
+                    "name": "InteractiveAgent",
+                    "role": "Assistant",
+                    "goal": "Help the user with their requests",
+                    "instructions": "You are a helpful AI assistant. Be concise and helpful.",
+                }
+                
+                if self.config.model:
+                    agent_kwargs["llm"] = self.config.model
+                
+                if self.config.memory:
+                    agent_kwargs["memory"] = True
+                
+                if self.config.tools:
+                    agent_kwargs["tools"] = self.config.tools
+                
+                self._agent = Agent(**agent_kwargs)
+                
+            except ImportError as e:
+                self.io.tool_error(f"Failed to import praisonaiagents: {e}")
+                raise
+        
+        return self._agent
+    
+    def _handle_command(self, command: str) -> bool:
+        """
+        Handle slash commands.
+        
+        Returns True if command was handled, False otherwise.
+        """
+        parts = command.split(maxsplit=1)
+        cmd = parts[0].lower().lstrip("/")
+        args = parts[1] if len(parts) > 1 else ""
+        
+        if cmd in ("exit", "quit", "q"):
+            self._running = False
+            self.io.info("Goodbye!")
+            return True
+        
+        elif cmd == "help":
+            self.io.print_help(DEFAULT_COMMANDS)
+            return True
+        
+        elif cmd == "clear":
+            self._conversation_history.clear()
+            self.io.success("Conversation cleared.")
+            return True
+        
+        elif cmd == "session":
+            if self._session_id:
+                self.io.info(f"Session: {self._session_id}")
+            else:
+                self.io.info("No active session")
+            return True
+        
+        elif cmd == "model":
+            if args:
+                self.config.model = args
+                self._agent = None  # Reset agent to use new model
+                self.io.success(f"Model changed to: {args}")
+            else:
+                self.io.info(f"Current model: {self.config.model}")
+            return True
+        
+        elif cmd == "cost":
+            self.io.info(f"Total tokens: {self._total_tokens}")
+            self.io.info(f"Estimated cost: ${self._total_cost:.4f}")
+            return True
+        
+        elif cmd == "history":
+            if not self._conversation_history:
+                self.io.info("No conversation history")
+            else:
+                for i, msg in enumerate(self._conversation_history[-10:], 1):
+                    role = msg.get("role", "unknown")
+                    content = msg.get("content", "")[:100]
+                    self.io.info(f"{i}. [{role}] {content}...")
+            return True
+        
+        elif cmd == "export":
+            filename = args or "conversation.txt"
+            try:
+                with open(filename, "w") as f:
+                    for msg in self._conversation_history:
+                        f.write(f"[{msg.get('role', 'unknown')}]\n")
+                        f.write(f"{msg.get('content', '')}\n\n")
+                self.io.success(f"Exported to {filename}")
+            except Exception as e:
+                self.io.tool_error(f"Export failed: {e}")
+            return True
+        
+        elif cmd == "compact":
+            self.io.config.pretty = not self.io.config.pretty
+            mode = "disabled" if self.io.config.pretty else "enabled"
+            self.io.info(f"Compact mode {mode}")
+            return True
+        
+        elif cmd == "multiline":
+            self.io.config.multiline_mode = not self.io.config.multiline_mode
+            mode = "enabled" if self.io.config.multiline_mode else "disabled"
+            self.io.info(f"Multiline mode {mode}")
+            return True
+        
+        else:
+            self.io.tool_warning(f"Unknown command: /{cmd}")
+            self.io.info("Type /help for available commands")
+            return True
+    
+    def _execute_prompt(self, prompt: str) -> Optional[str]:
+        """Execute a prompt and return the response."""
+        try:
+            agent = self._get_agent()
+            
+            # Add to history
+            self._conversation_history.append({
+                "role": "user",
+                "content": prompt
+            })
+            
+            # Execute
+            response = agent.start(prompt)
+            
+            # Add response to history
+            if response:
+                self._conversation_history.append({
+                    "role": "assistant", 
+                    "content": str(response)
+                })
+            
+            return str(response) if response else None
+            
+        except Exception as e:
+            logger.exception("Error executing prompt")
+            self.io.tool_error(f"Error: {e}")
+            return None
+    
+    def run(self) -> None:
+        """
+        Run the interactive REPL.
+        
+        This is the main entry point. It runs synchronously,
+        using prompt_toolkit for input (which must be sync).
+        """
+        self._running = True
+        
+        # Get tools count for display
+        tools_count = len(self.config.tools) if self.config.tools else 0
+        
+        # Generate session ID
+        import uuid
+        self._session_id = self.config.session_id or str(uuid.uuid4())
+        
+        # Print welcome
+        self.io.print_welcome(
+            model=self.config.model,
+            tools_count=tools_count,
+            session_id=self._session_id
+        )
+        
+        while self._running:
+            try:
+                # Get input (SYNC - this is the key!)
+                user_input = self.io.get_input("You: ").strip()
+                
+                if not user_input:
+                    continue
+                
+                # Handle slash commands
+                if user_input.startswith("/"):
+                    self._handle_command(user_input)
+                    continue
+                
+                # Show thinking indicator
+                self.io.print_assistant_start()
+                
+                # Execute prompt
+                response = self._execute_prompt(user_input)
+                
+                # Display response
+                if response:
+                    self.io.print_assistant_response(response)
+                
+            except KeyboardInterrupt:
+                self.io.info("\nUse /exit to quit")
+            except EOFError:
+                self._running = False
+                self.io.info("Goodbye!")
+            except Exception as e:
+                logger.exception("Error in REPL loop")
+                self.io.tool_error(f"Error: {e}")
+    
+    def run_single(self, prompt: str) -> Optional[str]:
+        """Run a single prompt and return the response."""
+        return self._execute_prompt(prompt)
+
+
+def start_interactive(
+    model: str = "gpt-4o-mini",
+    verbose: bool = False,
+    memory: bool = False,
+    tools: Optional[List] = None,
+    session_id: Optional[str] = None,
+    **kwargs
+) -> None:
+    """
+    Start the interactive REPL.
+    
+    This is the main entry point for `praisonai chat` and `praisonai` (no args).
+    """
+    config = REPLConfig(
+        model=model,
+        verbose=verbose,
+        memory=memory,
+        tools=tools,
+        session_id=session_id,
+    )
+    
+    repl = InteractiveREPL(config=config)
+    repl.run()

src/praisonai/praisonai/cli/interactive/split_tui.py

@@ -0,0 +1,456 @@
+"""
+PraisonAI Split-Pane TUI Application.
+
+Full-screen terminal UI with:
+- Fixed bottom input pane (always visible)
+- Scrollable top output pane
+- Direct response (no streaming - faster)
+- Activity indicators showing agent status
+
+Uses prompt_toolkit Application for true split-pane layout.
+"""
+
+import shutil
+import threading
+import time
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import List, Optional
+from datetime import datetime
+
+# ============================================================================
+# Branding - Import from unified source
+# ============================================================================
+
+from praisonai.cli.branding import get_logo, get_version
+
+
+# ============================================================================
+# Configuration
+# ============================================================================
+
+@dataclass
+class SplitTUIConfig:
+    """Configuration for the Split TUI application."""
+    model: str = "gpt-4o-mini"
+    show_logo: bool = True
+    show_status_bar: bool = True
+    session_id: Optional[str] = None
+    workspace: Optional[str] = None
+    history_file: Optional[str] = None
+
+
+# ============================================================================
+# Message Types
+# ============================================================================
+
+@dataclass
+class ChatMessage:
+    """A chat message."""
+    role: str  # "user", "assistant", "system", "status"
+    content: str
+    timestamp: datetime = field(default_factory=datetime.now)
+
+
+# ============================================================================
+# Split TUI Application
+# ============================================================================
+
+class SplitTUI:
+    """
+    Split-pane TUI with fixed bottom input and scrollable top output.
+    
+    Key features:
+    - Input box ALWAYS visible at bottom
+    - Output scrolls above independently
+    - Direct response (no streaming) for speed
+    - Activity indicators during processing
+    """
+    
+    def __init__(self, config: Optional[SplitTUIConfig] = None):
+        self.config = config or SplitTUIConfig()
+        self.messages: List[ChatMessage] = []
+        self._running = False
+        self._agent = None
+        self._processing = False
+        self._status_text = ""
+        
+        # Terminal size
+        self.term_width, self.term_height = shutil.get_terminal_size((80, 24))
+        
+        # Session ID
+        import uuid
+        self.session_id = self.config.session_id or str(uuid.uuid4())[:8]
+    
+    def _get_agent(self):
+        """Lazy-load the agent with verbose=False for no streaming."""
+        if self._agent is None:
+            try:
+                from praisonaiagents import Agent
+                
+                self._agent = Agent(
+                    name="Praison",
+                    role="AI Assistant",
+                    goal="Help the user with their requests",
+                    instructions="You are Praison AI, a helpful assistant. Be concise and helpful.",
+                    llm=self.config.model,
+                )
+            except ImportError as e:
+                raise RuntimeError(f"Failed to import praisonaiagents: {e}")
+        
+        return self._agent
+    
+    def _format_output(self) -> str:
+        """Format all messages for the output pane."""
+        lines = []
+        
+        # Logo
+        if self.config.show_logo and not self.messages:
+            logo = get_logo(self.term_width)
+            lines.append(logo.strip())
+            lines.append("")
+            lines.append(f"  v{get_version()} · Model: {self.config.model}")
+            lines.append("")
+            lines.append("  Type your message and press Enter. Use /help for commands.")
+            lines.append("")
+        
+        # Messages
+        for msg in self.messages:
+            if msg.role == "user":
+                lines.append(f"› {msg.content}")
+                lines.append("")
+            elif msg.role == "assistant":
+                lines.append(f"● {msg.content}")
+                lines.append("")
+            elif msg.role == "system":
+                lines.append(f"  {msg.content}")
+                lines.append("")
+            elif msg.role == "status":
+                lines.append(f"  ⏳ {msg.content}")
+        
+        # Processing indicator
+        if self._processing:
+            lines.append(f"  ⏳ {self._status_text}")
+        
+        return "\n".join(lines)
+    
+    def _handle_command(self, command: str) -> bool:
+        """Handle slash commands. Returns True if handled."""
+        parts = command.split(maxsplit=1)
+        cmd = parts[0].lower().lstrip("/")
+        args = parts[1] if len(parts) > 1 else ""
+        
+        if cmd in ("exit", "quit", "q"):
+            self._running = False
+            return True
+        
+        elif cmd == "help":
+            help_text = """Available Commands:
+  /help          Show this help
+  /exit, /quit   Exit Praison AI
+  /clear         Clear conversation
+  /model [name]  Show or change model
+  /new           Start new conversation"""
+            self.messages.append(ChatMessage(role="system", content=help_text))
+            return True
+        
+        elif cmd == "clear":
+            self.messages.clear()
+            return True
+        
+        elif cmd == "new":
+            self.messages.clear()
+            import uuid
+            self.session_id = str(uuid.uuid4())[:8]
+            self.messages.append(ChatMessage(role="system", content=f"New session: {self.session_id}"))
+            return True
+        
+        elif cmd == "model":
+            if args:
+                self.config.model = args
+                self._agent = None
+                self.messages.append(ChatMessage(role="system", content=f"Model changed to: {args}"))
+            else:
+                self.messages.append(ChatMessage(role="system", content=f"Current model: {self.config.model}"))
+            return True
+        
+        else:
+            self.messages.append(ChatMessage(role="system", content=f"Unknown command: /{cmd}. Use /help for available commands."))
+            return True
+    
+    def _execute_prompt(self, prompt: str) -> Optional[str]:
+        """Execute a prompt and return the response (direct, no streaming)."""
+        import sys
+        import io
+        
+        try:
+            agent = self._get_agent()
+            
+            # Suppress agent's Rich output by redirecting stdout/stderr
+            old_stdout = sys.stdout
+            old_stderr = sys.stderr
+            sys.stdout = io.StringIO()
+            sys.stderr = io.StringIO()
+            
+            try:
+                response = agent.start(prompt)
+            finally:
+                # Restore stdout/stderr
+                sys.stdout = old_stdout
+                sys.stderr = old_stderr
+            
+            return str(response) if response else None
+        except Exception as e:
+            return f"Error: {e}"
+    
+    def _run_with_status(self, prompt: str) -> Optional[str]:
+        """Run prompt with status updates."""
+        self._processing = True
+        self._status_text = "Praison AI is thinking..."
+        
+        # Execute in background
+        result = [None]
+        error = [None]
+        
+        def execute():
+            try:
+                result[0] = self._execute_prompt(prompt)
+            except Exception as e:
+                error[0] = str(e)
+        
+        thread = threading.Thread(target=execute)
+        thread.start()
+        
+        # Update status while waiting
+        start_time = time.time()
+        spinner = ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧", "⠇", "⠏"]
+        idx = 0
+        
+        while thread.is_alive():
+            elapsed = time.time() - start_time
+            self._status_text = f"{spinner[idx]} Praison AI is thinking... ({elapsed:.1f}s)"
+            idx = (idx + 1) % len(spinner)
+            time.sleep(0.1)
+        
+        thread.join()
+        self._processing = False
+        self._status_text = ""
+        
+        if error[0]:
+            return f"Error: {error[0]}"
+        return result[0]
+    
+    def run(self) -> None:
+        """Run the split-pane TUI."""
+        self._running = True
+        
+        # Try to use prompt_toolkit for better experience
+        try:
+            self._run_with_prompt_toolkit()
+        except ImportError:
+            # Fallback to simple mode
+            self._run_simple()
+    
+    def _run_simple(self) -> None:
+        """Simple fallback mode without prompt_toolkit."""
+        # Print welcome
+        print(self._format_output())
+        
+        while self._running:
+            try:
+                # Status bar
+                if self.config.show_status_bar:
+                    status = f" ? for help | {self.config.model} | Session: {self.session_id} "
+                    print(f"\033[90m{status}\033[0m")
+                
+                # Get input
+                user_input = input("› ").strip()
+                
+                if not user_input:
+                    continue
+                
+                # Handle commands
+                if user_input.startswith("/"):
+                    self._handle_command(user_input)
+                    # Print system messages
+                    for msg in self.messages:
+                        if msg.role == "system":
+                            print(f"  {msg.content}")
+                    self.messages = [m for m in self.messages if m.role != "system"]
+                    continue
+                
+                # Add user message
+                self.messages.append(ChatMessage(role="user", content=user_input))
+                
+                # Execute with status
+                print("  ⏳ Praison AI is thinking...")
+                response = self._execute_prompt(user_input)
+                
+                if response:
+                    self.messages.append(ChatMessage(role="assistant", content=response))
+                    print(f"\n● {response}\n")
+                
+            except KeyboardInterrupt:
+                print("\n  Use /exit to quit")
+            except EOFError:
+                self._running = False
+        
+        print("\n  Goodbye from Praison AI!")
+    
+    def _run_with_prompt_toolkit(self) -> None:
+        """Run with prompt_toolkit for split-pane layout."""
+        from prompt_toolkit import Application
+        from prompt_toolkit.buffer import Buffer
+        from prompt_toolkit.layout.containers import HSplit, VSplit, Window
+        from prompt_toolkit.layout.controls import BufferControl, FormattedTextControl
+        from prompt_toolkit.layout.layout import Layout
+        from prompt_toolkit.key_binding import KeyBindings
+        from prompt_toolkit.styles import Style
+        from prompt_toolkit.history import FileHistory
+        
+        # Setup history
+        history_file = self.config.history_file or str(Path.home() / ".praison" / "history")
+        Path(history_file).parent.mkdir(parents=True, exist_ok=True)
+        
+        # Output buffer (read-only, for displaying conversation)
+        output_text = [self._format_output()]
+        
+        def get_output_text():
+            return output_text[0]
+        
+        # Input buffer
+        input_buffer = Buffer(
+            multiline=False,
+            history=FileHistory(history_file),
+        )
+        
+        # Key bindings
+        kb = KeyBindings()
+        
+        @kb.add("enter")
+        def handle_enter(event):
+            user_input = input_buffer.text.strip()
+            if not user_input:
+                return
+            
+            input_buffer.reset()
+            
+            # Handle commands
+            if user_input.startswith("/"):
+                self._handle_command(user_input)
+                output_text[0] = self._format_output()
+                return
+            
+            # Add user message
+            self.messages.append(ChatMessage(role="user", content=user_input))
+            
+            # Show processing status
+            self.messages.append(ChatMessage(role="status", content="Praison AI is thinking..."))
+            output_text[0] = self._format_output()
+            
+            # Execute (this blocks but that's OK for direct response)
+            response = self._execute_prompt(user_input)
+            
+            # Remove status message
+            self.messages = [m for m in self.messages if m.role != "status"]
+            
+            if response:
+                self.messages.append(ChatMessage(role="assistant", content=response))
+            
+            output_text[0] = self._format_output()
+        
+        @kb.add("c-c")
+        def handle_ctrl_c(event):
+            input_buffer.reset()
+        
+        @kb.add("c-d")
+        def handle_ctrl_d(event):
+            if not input_buffer.text:
+                self._running = False
+                event.app.exit()
+        
+        @kb.add("c-q")
+        def handle_ctrl_q(event):
+            self._running = False
+            event.app.exit()
+        
+        # Style
+        style = Style.from_dict({
+            "output-area": "bg:#0d0d0d",
+            "input-area": "bg:#1a1a1a",
+            "status-bar": "bg:#333333 #888888",
+            "prompt": "#00aaff bold",
+        })
+        
+        # Status bar text
+        def get_status_bar():
+            left = "? for help"
+            center = self.config.model
+            right = f"Session: {self.session_id}"
+            return f" {left}  |  {center}  |  {right} "
+        
+        # Layout: Output (top, scrollable) + Status bar + Input (bottom, fixed)
+        layout = Layout(
+            HSplit([
+                # Output area (takes remaining space)
+                Window(
+                    content=FormattedTextControl(get_output_text),
+                    wrap_lines=True,
+                    style="class:output-area",
+                ),
+                # Status bar (1 line)
+                Window(
+                    content=FormattedTextControl(get_status_bar),
+                    height=1,
+                    style="class:status-bar",
+                ),
+                # Input area (fixed at bottom, 3 lines for visibility)
+                HSplit([
+                    Window(height=1),  # Spacer
+                    VSplit([
+                        Window(
+                            content=FormattedTextControl([("class:prompt", "› ")]),
+                            width=2,
+                        ),
+                        Window(
+                            content=BufferControl(buffer=input_buffer),
+                            wrap_lines=True,
+                        ),
+                    ]),
+                    Window(height=1),  # Spacer
+                ], height=3, style="class:input-area"),
+            ])
+        )
+        
+        # Create application
+        app = Application(
+            layout=layout,
+            key_bindings=kb,
+            style=style,
+            full_screen=True,
+            mouse_support=True,
+        )
+        
+        # Run
+        app.run()
+        
+        print("\n  Goodbye from Praison AI!")
+    
+    def run_single(self, prompt: str) -> Optional[str]:
+        """Run a single prompt (non-interactive)."""
+        return self._execute_prompt(prompt)
+
+
+def start_split_tui(
+    model: str = "gpt-4o-mini",
+    show_logo: bool = True,
+    **kwargs
+) -> None:
+    """Start the split-pane TUI application."""
+    config = SplitTUIConfig(
+        model=model,
+        show_logo=show_logo,
+    )
+    
+    tui = SplitTUI(config=config)
+    tui.run()

src/praisonai/praisonai/cli/interactive/tui_app.py

@@ -0,0 +1,457 @@
+"""
+PraisonAI Interactive TUI Application.
+
+Full-screen terminal UI with:
+- Sticky bottom input box (like Aider/Codex/Claude Code)
+- Scrollable output area above
+- ASCII art logo banner
+- Status bar with model/context info
+- Clean, spacious design inspired by Claude Code
+
+Uses prompt_toolkit for the full-screen application.
+"""
+
+import shutil
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Dict, List, Optional
+from datetime import datetime
+
+# Lazy imports for performance
+_PT_AVAILABLE = None
+_RICH_AVAILABLE = None
+
+
+def _check_prompt_toolkit():
+    global _PT_AVAILABLE
+    if _PT_AVAILABLE is None:
+        try:
+            import prompt_toolkit  # noqa: F401
+            _PT_AVAILABLE = True
+        except ImportError:
+            _PT_AVAILABLE = False
+    return _PT_AVAILABLE
+
+
+def _check_rich():
+    global _RICH_AVAILABLE
+    if _RICH_AVAILABLE is None:
+        try:
+            import rich  # noqa: F401
+            _RICH_AVAILABLE = True
+        except ImportError:
+            _RICH_AVAILABLE = False
+    return _RICH_AVAILABLE
+
+
+# ============================================================================
+# Branding - Import from unified source
+# ============================================================================
+
+from praisonai.cli.branding import get_logo, get_version
+
+
+# ============================================================================
+# Color Scheme
+# ============================================================================
+
+@dataclass
+class ColorScheme:
+    """Color scheme for the TUI."""
+    # Primary colors
+    primary: str = "#00aaff"      # Cyan-blue (PraisonAI brand)
+    secondary: str = "#888888"    # Gray
+    accent: str = "#00cc00"       # Green
+    
+    # Text colors
+    text: str = "#ffffff"         # White
+    text_dim: str = "#666666"     # Dim gray
+    text_muted: str = "#444444"   # Very dim
+    
+    # Status colors
+    success: str = "#00cc00"      # Green
+    warning: str = "#ffaa00"      # Orange
+    error: str = "#ff4444"        # Red
+    info: str = "#00aaff"         # Blue
+    
+    # UI elements
+    border: str = "#333333"       # Dark gray
+    input_bg: str = "#1a1a1a"     # Very dark
+    output_bg: str = "#0d0d0d"    # Almost black
+
+
+DEFAULT_COLORS = ColorScheme()
+
+
+# ============================================================================
+# TUI Configuration
+# ============================================================================
+
+@dataclass
+class TUIConfig:
+    """Configuration for the TUI application."""
+    # Model settings
+    model: str = "gpt-4o-mini"
+    
+    # Display settings
+    show_logo: bool = True
+    show_tips: bool = True
+    show_status_bar: bool = True
+    compact_mode: bool = False
+    
+    # Colors
+    colors: ColorScheme = field(default_factory=ColorScheme)
+    
+    # Input settings
+    multiline: bool = False
+    vi_mode: bool = False
+    
+    # History
+    history_file: Optional[str] = None
+    
+    # Session
+    session_id: Optional[str] = None
+    workspace: Optional[str] = None
+
+
+# ============================================================================
+# Message Types
+# ============================================================================
+
+@dataclass
+class Message:
+    """A chat message."""
+    role: str  # "user", "assistant", "system", "tool"
+    content: str
+    timestamp: datetime = field(default_factory=datetime.now)
+    metadata: Dict = field(default_factory=dict)
+
+
+# ============================================================================
+# TUI Application
+# ============================================================================
+
+class PraisonTUI:
+    """
+    Full-screen TUI application for PraisonAI.
+    
+    Features:
+    - Sticky bottom input box
+    - Scrollable output area
+    - ASCII art logo
+    - Status bar
+    - Clean, spacious design
+    """
+    
+    def __init__(self, config: Optional[TUIConfig] = None):
+        self.config = config or TUIConfig()
+        self.messages: List[Message] = []
+        self._running = False
+        self._agent = None
+        self._total_tokens = 0
+        self._total_cost = 0.0
+        
+        # Get terminal size
+        self.term_width, self.term_height = shutil.get_terminal_size((80, 24))
+        
+        # Generate session ID
+        import uuid
+        self.session_id = self.config.session_id or str(uuid.uuid4())[:8]
+    
+    def _get_agent(self):
+        """Lazy-load the agent."""
+        if self._agent is None:
+            try:
+                from praisonaiagents import Agent
+                
+                self._agent = Agent(
+                    name="PraisonAgent",
+                    role="Assistant",
+                    goal="Help the user with their requests",
+                    instructions="You are a helpful AI assistant. Be concise and helpful.",
+                    llm=self.config.model,
+                )
+            except ImportError as e:
+                raise RuntimeError(f"Failed to import praisonaiagents: {e}")
+        
+        return self._agent
+    
+    def _render_logo(self) -> str:
+        """Render the ASCII logo with colors."""
+        if not self.config.show_logo:
+            return ""
+        
+        logo = get_logo(self.term_width)
+        version = get_version()
+        
+        info_line = f"v{version} · Model: {self.config.model} · Session: {self.session_id}"
+        return logo.strip() + "\n" + info_line
+    
+    def _render_welcome(self) -> str:
+        """Render welcome message."""
+        lines = []
+        
+        # Logo
+        if self.config.show_logo:
+            lines.append(self._render_logo())
+            lines.append("")
+        
+        # Tips
+        if self.config.show_tips:
+            tips = [
+                "Type your message and press Enter to send",
+                "Use /help for commands, /exit to quit",
+                "Press ? for keyboard shortcuts",
+            ]
+            lines.append("  " + tips[0])
+        
+        lines.append("")
+        
+        return "\n".join(lines)
+    
+    def _render_status_bar(self) -> str:
+        """Render the status bar."""
+        if not self.config.show_status_bar:
+            return ""
+        
+        left = f"? for shortcuts"
+        center = f"{self.config.model}"
+        right = f"Session: {self.session_id}"
+        
+        # Calculate spacing
+        total_len = len(left) + len(center) + len(right)
+        available = self.term_width - total_len - 4
+        
+        if available > 0:
+            left_space = available // 2
+            right_space = available - left_space
+            return f" {left}{' ' * left_space}{center}{' ' * right_space}{right} "
+        else:
+            return f" {center} "
+    
+    def _format_message(self, msg: Message) -> str:
+        """Format a message for display."""
+        if msg.role == "user":
+            prefix = "› "
+            return f"{prefix}{msg.content}"
+        elif msg.role == "assistant":
+            prefix = "● "
+            return f"{prefix}{msg.content}"
+        elif msg.role == "tool":
+            prefix = "⚙ "
+            return f"{prefix}{msg.content}"
+        else:
+            return msg.content
+    
+    def _handle_command(self, command: str) -> bool:
+        """Handle slash commands. Returns True if handled."""
+        parts = command.split(maxsplit=1)
+        cmd = parts[0].lower().lstrip("/")
+        args = parts[1] if len(parts) > 1 else ""
+        
+        if cmd in ("exit", "quit", "q"):
+            self._running = False
+            return True
+        
+        elif cmd == "help":
+            help_text = """
+Available Commands:
+  /help          Show this help
+  /exit, /quit   Exit interactive mode
+  /clear         Clear conversation
+  /model [name]  Show or change model
+  /cost          Show token usage
+  /compact       Toggle compact mode
+  /multiline     Toggle multiline input
+"""
+            self.messages.append(Message(role="system", content=help_text.strip()))
+            return True
+        
+        elif cmd == "clear":
+            self.messages.clear()
+            return True
+        
+        elif cmd == "model":
+            if args:
+                self.config.model = args
+                self._agent = None
+                self.messages.append(Message(role="system", content=f"Model changed to: {args}"))
+            else:
+                self.messages.append(Message(role="system", content=f"Current model: {self.config.model}"))
+            return True
+        
+        elif cmd == "cost":
+            self.messages.append(Message(
+                role="system",
+                content=f"Tokens: {self._total_tokens} | Cost: ${self._total_cost:.4f}"
+            ))
+            return True
+        
+        elif cmd == "compact":
+            self.config.compact_mode = not self.config.compact_mode
+            mode = "enabled" if self.config.compact_mode else "disabled"
+            self.messages.append(Message(role="system", content=f"Compact mode {mode}"))
+            return True
+        
+        elif cmd == "multiline":
+            self.config.multiline = not self.config.multiline
+            mode = "enabled" if self.config.multiline else "disabled"
+            self.messages.append(Message(role="system", content=f"Multiline mode {mode}"))
+            return True
+        
+        else:
+            self.messages.append(Message(role="system", content=f"Unknown command: /{cmd}"))
+            return True
+    
+    def _execute_prompt(self, prompt: str) -> Optional[str]:
+        """Execute a prompt and return the response."""
+        try:
+            agent = self._get_agent()
+            response = agent.start(prompt)
+            return str(response) if response else None
+        except Exception as e:
+            return f"Error: {e}"
+    
+    def run(self) -> None:
+        """Run the TUI application."""
+        self._running = True
+        
+        # Print welcome
+        print(self._render_welcome())
+        
+        # Setup history
+        history_file = self.config.history_file or str(Path.home() / ".praison" / "history")
+        Path(history_file).parent.mkdir(parents=True, exist_ok=True)
+        
+        # Create prompt session
+        if _check_prompt_toolkit():
+            from prompt_toolkit import PromptSession
+            from prompt_toolkit.history import FileHistory
+            from prompt_toolkit.auto_suggest import AutoSuggestFromHistory
+            from prompt_toolkit.key_binding import KeyBindings
+            from prompt_toolkit.styles import Style
+            
+            # Key bindings
+            kb = KeyBindings()
+            
+            @kb.add("c-c")
+            def _(event):
+                event.current_buffer.reset()
+            
+            @kb.add("c-d")
+            def _(event):
+                if not event.current_buffer.text:
+                    event.app.exit()
+            
+            # Style
+            style = Style.from_dict({
+                "": self.config.colors.accent,
+                "prompt": self.config.colors.primary,
+            })
+            
+            session = PromptSession(
+                history=FileHistory(history_file),
+                auto_suggest=AutoSuggestFromHistory(),
+                key_bindings=kb,
+                style=style,
+                multiline=self.config.multiline,
+            )
+            
+            def get_input():
+                return session.prompt(
+                    [("class:prompt", "› ")],
+                    placeholder="Type your message or @path/to/file"
+                )
+        else:
+            def get_input():
+                return input("› ")
+        
+        # Main loop
+        while self._running:
+            try:
+                # Print status bar
+                if self.config.show_status_bar and not self.config.compact_mode:
+                    status = self._render_status_bar()
+                    if _check_rich():
+                        from rich.console import Console
+                        Console().print(f"[dim]{status}[/dim]")
+                    else:
+                        print(f"\033[90m{status}\033[0m")
+                
+                # Get input
+                user_input = get_input().strip()
+                
+                if not user_input:
+                    continue
+                
+                # Handle commands
+                if user_input.startswith("/"):
+                    self._handle_command(user_input)
+                    # Print system messages
+                    for msg in self.messages:
+                        if msg.role == "system":
+                            print(self._format_message(msg))
+                    self.messages = [m for m in self.messages if m.role != "system"]
+                    continue
+                
+                # Add user message
+                self.messages.append(Message(role="user", content=user_input))
+                
+                # Show thinking indicator
+                if _check_rich():
+                    from rich.console import Console
+                    Console().print("[dim]Thinking...[/dim]")
+                
+                # Execute
+                response = self._execute_prompt(user_input)
+                
+                if response:
+                    self.messages.append(Message(role="assistant", content=response))
+                    
+                    # Display response
+                    if _check_rich():
+                        from rich.console import Console
+                        from rich.markdown import Markdown
+                        from rich.panel import Panel
+                        
+                        console = Console()
+                        try:
+                            md = Markdown(response)
+                            console.print(Panel(md, border_style="cyan", padding=(0, 1)))
+                        except Exception:
+                            console.print(f"● {response}")
+                    else:
+                        print(f"● {response}")
+                
+                print()  # Spacing
+                
+            except KeyboardInterrupt:
+                print("\nUse /exit to quit")
+            except EOFError:
+                self._running = False
+                print("\nGoodbye!")
+            except Exception as e:
+                print(f"Error: {e}")
+        
+        print("Goodbye!")
+    
+    def run_single(self, prompt: str) -> Optional[str]:
+        """Run a single prompt."""
+        return self._execute_prompt(prompt)
+
+
+def start_tui(
+    model: str = "gpt-4o-mini",
+    show_logo: bool = True,
+    compact: bool = False,
+    **kwargs
+) -> None:
+    """Start the TUI application."""
+    config = TUIConfig(
+        model=model,
+        show_logo=show_logo,
+        compact_mode=compact,
+    )
+    
+    tui = PraisonTUI(config=config)
+    tui.run()

src/praisonai/praisonai/deploy.py

@@ -57,7 +57,7 @@ class CloudDeployer:
             file.write("FROM python:3.11-slim\n")
             file.write("WORKDIR /app\n")
             file.write("COPY . .\n")
-            file.write("RUN pip install flask praisonai==3.7.3 gunicorn markdown\n")
+            file.write("RUN pip install flask praisonai==3.8.0 gunicorn markdown\n")
             file.write("EXPOSE 8080\n")
             file.write('CMD ["gunicorn", "-b", "0.0.0.0:8080", "api:app"]\n')
             

src/praisonai/praisonai/standardise/ai_generator.py

@@ -358,6 +358,69 @@ class AIGenerator:
                 pass
         return self._fast_context_handler
     
+    def _detect_feature_capabilities(self, slug_str: str, source_info: Dict[str, Any]) -> Dict[str, Any]:
+        """Detect feature capabilities from source analysis."""
+        caps = {
+            "supports_bool": True,
+            "supports_list": False,
+            "supports_dict": False,
+            "supports_presets": [],
+            "supports_url": False,
+            "usage_forms": [],
+            "decorator": None,
+            "built_in_items": [],
+        }
+        
+        # Feature-specific detection based on slug
+        if slug_str == "tools":
+            caps["supports_list"] = True
+            caps["supports_bool"] = False
+            caps["decorator"] = "@tool"
+            caps["built_in_items"] = ["duckduckgo", "wikipedia_tools", "read_file", "execute_command"]
+            caps["usage_forms"] = [
+                "- List of tools: `tools=[my_func, another_func]`",
+                "- Using @tool decorator to create custom tools",
+                "- Using built-in tools: duckduckgo, wikipedia_tools, etc.",
+            ]
+        elif slug_str == "memory":
+            caps["supports_dict"] = True
+            caps["supports_presets"] = ["sqlite", "redis", "postgres", "qdrant", "chroma"]
+            caps["supports_url"] = True
+            caps["usage_forms"] = [
+                "- Boolean: `memory=True`",
+                "- Preset: `memory=\"sqlite\"` or `memory=\"redis\"`",
+                "- URL: `memory=\"redis://localhost:6379\"`",
+                "- Dict: `memory={\"provider\": \"qdrant\"}`",
+            ]
+        elif slug_str == "knowledge":
+            caps["supports_dict"] = True
+            caps["supports_list"] = True
+            caps["usage_forms"] = [
+                "- Boolean: `knowledge=True`",
+                "- List of sources: `knowledge=[\"file.pdf\", \"doc.txt\"]`",
+                "- Dict config: `knowledge={\"sources\": [...]}`",
+            ]
+        elif slug_str == "guardrails":
+            caps["supports_list"] = True
+            caps["decorator"] = "guardrail function"
+            caps["usage_forms"] = [
+                "- Boolean: `guardrails=True`",
+                "- List of functions: `guardrails=[my_guardrail_func]`",
+                "- Custom guardrail function returning GuardrailResult",
+            ]
+        elif slug_str == "streaming":
+            caps["supports_bool"] = True
+            caps["usage_forms"] = [
+                "- Boolean: `output=\"stream\"` or iterate over agent.start()",
+                "- Streaming with callbacks",
+            ]
+        else:
+            # Generic detection from source
+            caps["usage_forms"] = [f"- Boolean: `{slug_str}=True`"]
+        
+        logger.debug(f"Feature '{slug_str}' capabilities: {caps}")
+        return caps
+    
     def _get_agent(self):
         """Lazy-load the generation agent."""
         if self._agent is None:
@@ -591,14 +654,19 @@ ISSUES: List any issues found (or "None")
         context = {}
         slug_str = slug.normalised
         
+        logger.debug(f"Gathering context for feature '{slug_str}', artifact '{artifact_type.value}'")
+        
         # Use FastContext to find relevant code in SDK
         fast_context = self._get_fast_context_handler()
         if fast_context and self.sdk_root:
+            logger.debug(f"FastContext available, searching in {self.sdk_root}")
             try:
                 # Search for feature-related code
                 query = f"Find {slug_str} implementation, usage examples, and API"
                 matches = fast_context.search_context(query, str(self.sdk_root), use_llm_keywords=False)
                 
+                logger.debug(f"FastContext found {len(matches)} matches for '{query}'")
+                
                 # Read matched files
                 for match in matches[:3]:
                     file_path = self.sdk_root / match.get('file', '')
@@ -606,10 +674,11 @@ ISSUES: List any issues found (or "None")
                         try:
                             content = file_path.read_text(encoding="utf-8")[:2000]
                             context[f"sdk_{file_path.stem}"] = content
+                            logger.debug(f"Added SDK context: {file_path.name} ({len(content)} chars)")
                         except Exception:
                             pass
-            except Exception:
-                pass
+            except Exception as e:
+                logger.debug(f"FastContext search failed: {e}")
         
         # Get SDK module info (fallback)
         if self.sdk_root and not context:
@@ -653,6 +722,7 @@ ISSUES: List any issues found (or "None")
                 if doc_path.exists():
                     context[f"existing_{doc_type}"] = doc_path.read_text(encoding="utf-8")[:2000]
         
+        logger.debug(f"Context gathered for '{slug_str}': {list(context.keys())}")
         return context
     
     def _read_sdk_module(self, sdk_path: Path) -> str:
@@ -866,17 +936,26 @@ print(result)
             "",
         ]
         
-        # Add usage forms guidance (common patterns for all features)
-        param_name = slug_str.replace("-", "_")
-        prompt_parts.extend([
-            "## USAGE FORMS TO DEMONSTRATE (MUST cover ALL these progressively):",
-            f"- Boolean: `{param_name}=True` - Enable with defaults",
-            f"- String preset: `{param_name}=\"sqlite\"` or `{param_name}=\"redis\"` - Use predefined backend",
-            f"- URL: `{param_name}=\"redis://localhost:6379\"` - Backend-specific connection",
-            f"- Dict config: `{param_name}={{\"backend\": \"postgres\", \"user_id\": \"u1\"}}` - Custom config",
-            f"- Array + overrides: `{param_name}=[\"redis\", {{\"user_id\": \"u1\"}}]` - Preset + customization",
-            "",
-        ])
+        # Get dynamic feature capabilities - NO hardcoded usage forms
+        capabilities = self._detect_feature_capabilities(slug_str, source_info)
+        
+        # Add ONLY the usage forms this feature actually supports
+        prompt_parts.append("## USAGE FORMS TO DEMONSTRATE (cover these progressively):")
+        for usage_form in capabilities["usage_forms"]:
+            prompt_parts.append(usage_form)
+        
+        # Add decorator info if applicable
+        if capabilities.get("decorator"):
+            prompt_parts.append(f"- MUST demonstrate the {capabilities['decorator']} pattern")
+        
+        # Add built-in items if applicable
+        if capabilities.get("built_in_items"):
+            items = ", ".join(capabilities["built_in_items"][:4])
+            prompt_parts.append(f"- Built-in options to show: {items}")
+        
+        prompt_parts.append("")
+        
+        logger.debug(f"Feature '{slug_str}' usage forms: {capabilities['usage_forms']}")
         
         # Add dynamically extracted parameters if available
         main_class = source_info.get("main_class")

src/praisonai/praisonai/version.py

@@ -1 +1 @@
-__version__ = "3.7.3"
+__version__ = "3.8.0"

src/praisonai/pyproject.toml

@@ -12,7 +12,7 @@ dependencies = [
     "rich>=13.7",
     "markdown>=3.5",
     "pyparsing>=3.0.0",
-    "praisonaiagents>=0.11.2",
+    "praisonaiagents>=0.11.6",
     "python-dotenv>=0.19.0",
     "instructor>=1.3.3",
     "PyYAML>=6.0",

src/praisonai/tests/live/interactive/__init__.py

@@ -0,0 +1,11 @@
+"""Live interactive mode smoke tests.
+
+These tests verify PraisonAI's interactive mode using real API calls.
+Tests use HeadlessInteractiveCore to execute prompts through the same
+pipeline as the interactive TUI.
+
+Environment variables:
+- PRAISONAI_LIVE_INTERACTIVE=1  - Enable live interactive tests
+- OPENAI_API_KEY                - Required for live tests
+- PRAISONAI_LIVE_MODEL          - Override model (default: gpt-4o-mini)
+"""

src/praisonai/tests/live/interactive/advanced/__init__.py

@@ -0,0 +1,40 @@
+"""
+Advanced Agent-Centric Test Scenarios.
+
+These scenarios test the agent's ability to AUTONOMOUSLY complete complex tasks
+using ONLY a single text prompt. The agent must figure out the approach itself.
+
+Key Principles:
+- ONE prompt per scenario (no step-by-step instructions)
+- Prompt describes WHAT to achieve, not HOW
+- Agent uses tools autonomously to complete the task
+- Verification checks the END RESULT, not the process
+"""
+
+import os
+
+
+def is_advanced_tests_enabled() -> bool:
+    """Check if advanced tests are enabled."""
+    return os.environ.get("PRAISONAI_LIVE_INTERACTIVE", "0") == "1"
+
+
+def has_api_key() -> bool:
+    """Check if API key is available."""
+    return bool(os.environ.get("OPENAI_API_KEY"))
+
+
+def can_run_advanced_tests() -> tuple:
+    """Check if advanced tests can run."""
+    if not is_advanced_tests_enabled():
+        return False, "PRAISONAI_LIVE_INTERACTIVE=1 not set"
+    if not has_api_key():
+        return False, "OPENAI_API_KEY not set"
+    return True, "OK"
+
+
+__all__ = [
+    "is_advanced_tests_enabled",
+    "has_api_key", 
+    "can_run_advanced_tests",
+]

src/praisonai/tests/live/interactive/advanced/scenarios.py

@@ -0,0 +1,245 @@
+"""
+Advanced Agent-Centric Test Scenarios.
+
+5 complex scenarios where the agent must AUTONOMOUSLY figure out how to complete
+the task using ONLY a single text prompt. No step-by-step instructions.
+
+Scenarios:
+- ADV_01: Bug Fix & Testing - Find bug, write tests, fix it
+- ADV_02: Data Pipeline - Create JSON to CSV transformer
+- ADV_03: API Client - Build REST client with error handling
+- ADV_04: Documentation - Analyze code and generate docs
+- ADV_05: Package Creation - Create complete Python package
+"""
+
+from typing import List
+
+from ..runner import InteractiveScenario
+
+
+# =============================================================================
+# ADV_01: BUG FIX & TESTING
+# =============================================================================
+
+SCENARIO_ADV_01 = InteractiveScenario(
+    id="adv_01",
+    name="Bug Fix & Testing",
+    description="Agent must find a bug, write a test that exposes it, fix it, and verify",
+    prompts=[
+        """I have a Python file called calculator.py with the following buggy code:
+
+```python
+def add(a, b):
+    return a + b
+
+def subtract(a, b):
+    return a + b  # BUG: should be a - b
+
+def multiply(a, b):
+    return a * b
+
+def divide(a, b):
+    return a / b
+```
+
+Your task:
+1. Create this calculator.py file
+2. Create a test file test_calculator.py with tests for ALL functions
+3. Run the tests to find which one fails
+4. Fix the bug in calculator.py
+5. Run tests again to verify the fix works
+
+Show me the final working code and test results."""
+    ],
+    expected_tools=["acp_create_file"],
+    expected_files={
+        "calculator.py": r"def subtract.*return a - b",
+        "test_calculator.py": r"def test_|import.*calculator",
+    },
+    timeout=180,
+)
+
+
+# =============================================================================
+# ADV_02: DATA PIPELINE
+# =============================================================================
+
+SCENARIO_ADV_02 = InteractiveScenario(
+    id="adv_02",
+    name="Data Pipeline",
+    description="Agent must create a data transformation pipeline",
+    prompts=[
+        """Create a complete data processing pipeline that:
+
+1. Creates a sample data file called 'users.json' with this content:
+   [
+     {"name": "Alice", "age": 30, "city": "NYC"},
+     {"name": "Bob", "age": 25, "city": "LA"},
+     {"name": "Charlie", "age": 35, "city": "Chicago"}
+   ]
+
+2. Creates a Python script called 'transform.py' that:
+   - Reads the JSON file
+   - Filters users older than 28
+   - Converts to CSV format
+   - Saves to 'output.csv'
+
+3. Run the script to generate output.csv
+
+4. Show me the contents of output.csv
+
+The agent should figure out the best approach to accomplish this."""
+    ],
+    expected_tools=["acp_create_file"],
+    expected_files={
+        "users.json": r"Alice|Bob|Charlie",
+        "transform.py": r"import json|csv|open",
+    },
+    expected_response=r"Alice|Charlie|csv|output",
+    timeout=180,
+)
+
+
+# =============================================================================
+# ADV_03: API CLIENT
+# =============================================================================
+
+SCENARIO_ADV_03 = InteractiveScenario(
+    id="adv_03",
+    name="API Client",
+    description="Agent must build a REST API client with proper error handling",
+    prompts=[
+        """Create a Python module for making HTTP requests with proper error handling.
+
+Requirements:
+1. Create a file called 'api_client.py' with a class called 'APIClient' that has:
+   - __init__(self, base_url) - stores the base URL
+   - get(self, endpoint) - makes GET request, returns JSON response
+   - post(self, endpoint, data) - makes POST request with JSON body
+   - Proper error handling for network errors and HTTP errors
+   - A timeout parameter (default 30 seconds)
+
+2. Create a file called 'test_api_client.py' with unit tests that:
+   - Test the class can be instantiated
+   - Test error handling works (mock or check exception types)
+
+3. Show me the complete implementation
+
+The code should be production-quality with docstrings and type hints."""
+    ],
+    expected_tools=["acp_create_file"],
+    expected_files={
+        "api_client.py": r"class APIClient|def get|def post|except|timeout",
+        "test_api_client.py": r"def test_|APIClient",
+    },
+    timeout=180,
+)
+
+
+# =============================================================================
+# ADV_04: DOCUMENTATION GENERATOR
+# =============================================================================
+
+SCENARIO_ADV_04 = InteractiveScenario(
+    id="adv_04",
+    name="Documentation Generator",
+    description="Agent must analyze code and generate comprehensive documentation",
+    prompts=[
+        """I need you to create a Python module and then generate documentation for it.
+
+Step 1: Create a file called 'utils.py' with these utility functions:
+- format_date(date_obj) - formats a date to 'YYYY-MM-DD'
+- parse_csv_line(line) - splits a CSV line handling quoted fields
+- slugify(text) - converts text to URL-friendly slug
+- truncate(text, max_length) - truncates text with ellipsis
+
+Each function should have proper implementation (not just pass).
+
+Step 2: Create a file called 'README.md' that documents:
+- Module overview
+- Each function with description, parameters, return value, and example usage
+- Installation/usage instructions
+
+Make the documentation clear and professional."""
+    ],
+    expected_tools=["acp_create_file"],
+    expected_files={
+        "utils.py": r"def format_date|def parse_csv|def slugify|def truncate",
+        "README.md": r"format_date|parse_csv|slugify|truncate|##|Usage",
+    },
+    timeout=180,
+)
+
+
+# =============================================================================
+# ADV_05: COMPLETE PACKAGE
+# =============================================================================
+
+SCENARIO_ADV_05 = InteractiveScenario(
+    id="adv_05",
+    name="Complete Package",
+    description="Agent must create a complete Python package structure",
+    prompts=[
+        """Create a complete Python package called 'mathtools' with proper structure.
+
+The package should include:
+
+1. Package structure:
+   - mathtools/__init__.py (exports main functions)
+   - mathtools/operations.py (basic math: add, subtract, multiply, divide)
+   - mathtools/statistics.py (mean, median, mode functions)
+   - tests/test_operations.py (tests for operations)
+   - tests/test_statistics.py (tests for statistics)
+   - pyproject.toml (modern Python packaging)
+   - README.md (package documentation)
+
+2. All functions should have:
+   - Type hints
+   - Docstrings
+   - Proper implementation (not stubs)
+
+3. Tests should be runnable with pytest
+
+Create all files and show me the package structure."""
+    ],
+    expected_tools=["acp_create_file"],
+    expected_files={
+        "mathtools/__init__.py": r"from.*import|__all__",
+        "mathtools/operations.py": r"def add|def subtract|def multiply|def divide",
+        "mathtools/statistics.py": r"def mean|def median",
+        "pyproject.toml": r"name.*=.*mathtools|project",
+    },
+    timeout=240,
+)
+
+
+# =============================================================================
+# ALL SCENARIOS
+# =============================================================================
+
+ALL_ADVANCED_SCENARIOS: List[InteractiveScenario] = [
+    SCENARIO_ADV_01,
+    SCENARIO_ADV_02,
+    SCENARIO_ADV_03,
+    SCENARIO_ADV_04,
+    SCENARIO_ADV_05,
+]
+
+
+def get_scenario_by_id(scenario_id: str) -> InteractiveScenario:
+    """Get a scenario by its ID."""
+    for s in ALL_ADVANCED_SCENARIOS:
+        if s.id == scenario_id:
+            return s
+    raise ValueError(f"Scenario {scenario_id} not found")
+
+
+__all__ = [
+    "SCENARIO_ADV_01",
+    "SCENARIO_ADV_02",
+    "SCENARIO_ADV_03",
+    "SCENARIO_ADV_04",
+    "SCENARIO_ADV_05",
+    "ALL_ADVANCED_SCENARIOS",
+    "get_scenario_by_id",
+]

src/praisonai/tests/live/interactive/advanced/test_advanced.py

@@ -0,0 +1,64 @@
+"""
+Advanced Agent-Centric Test Suite.
+
+Tests the agent's ability to autonomously complete complex tasks
+using ONLY a single text prompt per scenario.
+
+Run with: PRAISONAI_LIVE_INTERACTIVE=1 pytest tests/live/interactive/advanced/test_advanced.py -v
+"""
+
+import pytest
+
+from ..runner import LiveInteractiveRunner
+from . import can_run_advanced_tests
+from .scenarios import ALL_ADVANCED_SCENARIOS
+
+
+pytestmark = pytest.mark.skipif(
+    not can_run_advanced_tests()[0],
+    reason=f"Advanced tests disabled: {can_run_advanced_tests()[1]}"
+)
+
+
+@pytest.fixture(scope="module")
+def runner():
+    """Create a shared runner for all tests."""
+    return LiveInteractiveRunner(verbose=True)
+
+
+class TestAdvancedScenarios:
+    """Advanced agent-centric test scenarios."""
+
+    @pytest.mark.parametrize("scenario", ALL_ADVANCED_SCENARIOS, ids=lambda s: s.id)
+    def test_scenario(self, scenario, runner):
+        """Run a single advanced scenario with retry."""
+        result = runner._run_with_retry(scenario)
+        
+        if not result.passed:
+            print(f"\n{'='*60}")
+            print(f"SCENARIO FAILED: {scenario.id} - {scenario.name}")
+            print(f"{'='*60}")
+            print(f"Description: {scenario.description}")
+            print(f"Prompt: {scenario.prompts[0][:200]}...")
+            print(f"Attempts: {result.attempts}")
+            print(f"Assertions: {result.assertions}")
+            print(f"Tool calls: {result.tool_calls}")
+            print(f"Error: {result.error}")
+            print(f"\nResponse (last 500 chars):\n{result.response[-500:] if result.response else 'None'}")
+            print(f"{'='*60}\n")
+        
+        assert result.passed, f"Scenario {scenario.id} failed after {result.attempts} attempts: {result.error}"
+
+
+def run_all_and_report():
+    """Run all scenarios and print report (for direct execution)."""
+    runner = LiveInteractiveRunner(verbose=True)
+    summary = runner.run_all(ALL_ADVANCED_SCENARIOS)
+    summary.print_summary()
+    return summary.pass_rate == 1.0
+
+
+if __name__ == "__main__":
+    import sys
+    success = run_all_and_report()
+    sys.exit(0 if success else 1)

src/praisonai/tests/live/interactive/runner.py

@@ -0,0 +1,387 @@
+"""Live interactive mode test runner.
+
+This runner executes tests using HeadlessInteractiveCore - the same pipeline
+as the interactive TUI. Tests run with real API calls and retry on failure.
+
+Environment variables:
+- PRAISONAI_LIVE_INTERACTIVE=1  - Enable live interactive tests
+- OPENAI_API_KEY                - Required for live tests
+- PRAISONAI_LIVE_MODEL          - Override model (default: gpt-4o-mini)
+- PRAISON_APPROVAL_MODE         - Auto-set to 'auto' for non-interactive
+"""
+
+import json
+import os
+import shutil
+import tempfile
+import time
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+DEFAULT_MODEL = os.environ.get("PRAISONAI_LIVE_MODEL", "gpt-4o-mini")
+MAX_RETRIES = 3
+DEFAULT_TIMEOUT = 60
+
+
+def is_live_interactive_enabled() -> bool:
+    """Check if live interactive tests are enabled."""
+    return os.environ.get("PRAISONAI_LIVE_INTERACTIVE", "0") == "1"
+
+
+def has_api_key() -> bool:
+    """Check if OpenAI API key is available."""
+    return bool(os.environ.get("OPENAI_API_KEY"))
+
+
+@dataclass
+class ScenarioResult:
+    """Result of a scenario execution."""
+    scenario_id: str
+    name: str
+    passed: bool
+    attempts: int
+    duration: float
+    response: str = ""
+    tool_calls: List[str] = field(default_factory=list)
+    error: Optional[str] = None
+    assertions: Dict[str, bool] = field(default_factory=dict)
+    artifacts_path: Optional[str] = None
+    
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "scenario_id": self.scenario_id,
+            "name": self.name,
+            "passed": self.passed,
+            "attempts": self.attempts,
+            "duration": self.duration,
+            "response": self.response[:500] if self.response else "",
+            "tool_calls": self.tool_calls,
+            "error": self.error,
+            "assertions": self.assertions,
+            "artifacts_path": self.artifacts_path,
+        }
+
+
+@dataclass
+class InteractiveScenario:
+    """A test scenario for interactive mode verification."""
+    id: str
+    name: str
+    description: str
+    prompts: List[str]
+    expected_tools: List[str] = field(default_factory=list)
+    forbidden_tools: List[str] = field(default_factory=list)
+    expected_files: Dict[str, str] = field(default_factory=dict)
+    expected_response: str = ""
+    workspace_fixture: str = "empty"
+    max_retries: int = MAX_RETRIES
+    timeout: int = DEFAULT_TIMEOUT
+    model: str = ""
+
+
+@dataclass
+class RunSummary:
+    """Summary of all scenario runs."""
+    total: int = 0
+    passed: int = 0
+    failed: int = 0
+    duration: float = 0.0
+    results: List[ScenarioResult] = field(default_factory=list)
+    
+    @property
+    def pass_rate(self) -> float:
+        return self.passed / self.total if self.total > 0 else 0.0
+    
+    def to_dict(self) -> Dict[str, Any]:
+        return {
+            "total": self.total,
+            "passed": self.passed,
+            "failed": self.failed,
+            "pass_rate": self.pass_rate,
+            "duration": self.duration,
+            "results": [r.to_dict() for r in self.results],
+        }
+    
+    def print_summary(self) -> None:
+        """Print summary to console."""
+        try:
+            from rich.console import Console
+            from rich.table import Table
+            
+            console = Console()
+            
+            table = Table(title="Live Interactive Test Results")
+            table.add_column("ID", style="cyan")
+            table.add_column("Name", style="white")
+            table.add_column("Status")
+            table.add_column("Attempts", justify="right")
+            table.add_column("Duration", justify="right")
+            
+            for r in self.results:
+                status = "[green]✓ PASS[/green]" if r.passed else "[red]✗ FAIL[/red]"
+                table.add_row(
+                    r.scenario_id,
+                    r.name[:30],
+                    status,
+                    str(r.attempts),
+                    f"{r.duration:.2f}s",
+                )
+            
+            console.print(table)
+            console.print()
+            console.print(f"[bold]Total:[/bold] {self.total} | "
+                         f"[green]Passed:[/green] {self.passed} | "
+                         f"[red]Failed:[/red] {self.failed} | "
+                         f"[bold]Pass Rate:[/bold] {self.pass_rate:.1%}")
+            console.print(f"[bold]Total Duration:[/bold] {self.duration:.2f}s")
+            
+        except ImportError:
+            print("\n=== Live Interactive Test Results ===")
+            for r in self.results:
+                status = "✓" if r.passed else "✗"
+                print(f"  {status} {r.scenario_id}: {r.name} ({r.attempts} attempts, {r.duration:.2f}s)")
+            print(f"\nTotal: {self.total} | Passed: {self.passed} | Failed: {self.failed}")
+            print(f"Pass Rate: {self.pass_rate:.1%} | Duration: {self.duration:.2f}s")
+
+
+class LiveInteractiveRunner:
+    """
+    Live test runner for interactive mode using HeadlessInteractiveCore.
+    
+    Runs scenarios with retry logic until they pass or max retries reached.
+    """
+    
+    def __init__(
+        self,
+        model: str = "",
+        artifacts_dir: Optional[Path] = None,
+        keep_artifacts: bool = True,
+        verbose: bool = True,
+    ):
+        self.model = model or DEFAULT_MODEL
+        self.artifacts_dir = artifacts_dir or Path(tempfile.mkdtemp(prefix="praison_live_"))
+        self.keep_artifacts = keep_artifacts
+        self.verbose = verbose
+        self.results: List[ScenarioResult] = []
+        
+        # Ensure approval mode is auto for non-interactive
+        os.environ["PRAISON_APPROVAL_MODE"] = "auto"
+    
+    def run_all(self, scenarios: List[InteractiveScenario]) -> RunSummary:
+        """
+        Run all scenarios, retrying failures until satisfied.
+        
+        Args:
+            scenarios: List of scenarios to run
+            
+        Returns:
+            RunSummary with all results
+        """
+        start_time = time.time()
+        self.results = []
+        
+        if self.verbose:
+            print(f"\n{'='*60}")
+            print(f"Running {len(scenarios)} live interactive tests")
+            print(f"Model: {self.model}")
+            print(f"Artifacts: {self.artifacts_dir}")
+            print(f"{'='*60}\n")
+        
+        for i, scenario in enumerate(scenarios):
+            if self.verbose:
+                print(f"\n[{i+1}/{len(scenarios)}] {scenario.id}: {scenario.name}")
+                print(f"  Description: {scenario.description}")
+            
+            result = self._run_with_retry(scenario)
+            self.results.append(result)
+            
+            if self.verbose:
+                status = "✓ PASS" if result.passed else "✗ FAIL"
+                print(f"  Result: {status} (attempts: {result.attempts}, {result.duration:.2f}s)")
+                if result.error:
+                    print(f"  Error: {result.error}")
+        
+        summary = RunSummary(
+            total=len(self.results),
+            passed=sum(1 for r in self.results if r.passed),
+            failed=sum(1 for r in self.results if not r.passed),
+            duration=time.time() - start_time,
+            results=self.results,
+        )
+        
+        # Save summary
+        summary_path = self.artifacts_dir / "summary.json"
+        with open(summary_path, "w") as f:
+            json.dump(summary.to_dict(), f, indent=2)
+        
+        return summary
+    
+    def _run_with_retry(self, scenario: InteractiveScenario) -> ScenarioResult:
+        """Run a scenario with retry logic."""
+        start_time = time.time()
+        last_error = None
+        last_response = ""
+        last_tool_calls = []
+        last_assertions = {}
+        
+        for attempt in range(1, scenario.max_retries + 1):
+            if self.verbose and attempt > 1:
+                print(f"  Retry {attempt}/{scenario.max_retries}...")
+            
+            try:
+                result = self._run_single(scenario, attempt)
+                
+                if result.passed:
+                    return result
+                
+                last_error = result.error
+                last_response = result.response
+                last_tool_calls = result.tool_calls
+                last_assertions = result.assertions
+                
+            except Exception as e:
+                last_error = str(e)
+                if self.verbose:
+                    print(f"  Attempt {attempt} error: {e}")
+        
+        # All retries exhausted
+        return ScenarioResult(
+            scenario_id=scenario.id,
+            name=scenario.name,
+            passed=False,
+            attempts=scenario.max_retries,
+            duration=time.time() - start_time,
+            response=last_response,
+            tool_calls=last_tool_calls,
+            error=last_error or "Max retries exhausted",
+            assertions=last_assertions,
+            artifacts_path=str(self.artifacts_dir / scenario.id),
+        )
+    
+    def _run_single(self, scenario: InteractiveScenario, attempt: int) -> ScenarioResult:
+        """Run a single attempt of a scenario."""
+        start_time = time.time()
+        
+        # Create workspace
+        workspace = Path(tempfile.mkdtemp(prefix=f"praison_{scenario.id}_"))
+        artifacts_path = self.artifacts_dir / scenario.id / f"attempt_{attempt}"
+        
+        try:
+            # Import harness (lazy)
+            from praisonai.cli.features.interactive_test_harness import (
+                InteractiveTestHarness,
+            )
+            
+            # Create harness
+            harness = InteractiveTestHarness(
+                workspace=workspace,
+                artifacts_dir=artifacts_path,
+                keep_workspace=self.keep_artifacts,
+            )
+            
+            # Setup workspace
+            harness.setup_workspace(scenario.workspace_fixture)
+            
+            # Run prompts
+            exec_result = harness.run(
+                prompts=scenario.prompts,
+                model=scenario.model or self.model,
+                approval_mode="auto",
+            )
+            
+            duration = time.time() - start_time
+            
+            # Get response
+            response = "\n".join(exec_result.responses) if exec_result.responses else ""
+            
+            # Verify assertions
+            assertions = {}
+            
+            # Tool call assertions
+            tool_result = harness.verify_tool_calls(
+                expected_tools=scenario.expected_tools,
+                forbidden_tools=scenario.forbidden_tools,
+            )
+            assertions["tools"] = tool_result["passed"]
+            
+            # File assertions
+            if scenario.expected_files:
+                file_results = harness.verify_files(scenario.expected_files)
+                assertions["files"] = all(file_results.values())
+            else:
+                assertions["files"] = True
+            
+            # Response assertion
+            if scenario.expected_response:
+                assertions["response"] = harness.verify_response(response, scenario.expected_response)
+            else:
+                assertions["response"] = True
+            
+            # Execution success
+            assertions["execution"] = exec_result.success
+            
+            # Determine overall pass
+            all_passed = all(assertions.values())
+            
+            # Snapshot workspace if keeping artifacts
+            if self.keep_artifacts:
+                harness.snapshot_workspace()
+            
+            # Save artifacts
+            harness.save_artifacts(f"attempt_{attempt}")
+            
+            return ScenarioResult(
+                scenario_id=scenario.id,
+                name=scenario.name,
+                passed=all_passed,
+                attempts=attempt,
+                duration=duration,
+                response=response,
+                tool_calls=harness._executor.get_tools_called() if harness._executor else [],
+                error=exec_result.error if not exec_result.success else None,
+                assertions=assertions,
+                artifacts_path=str(artifacts_path),
+            )
+            
+        except Exception as e:
+            return ScenarioResult(
+                scenario_id=scenario.id,
+                name=scenario.name,
+                passed=False,
+                attempts=attempt,
+                duration=time.time() - start_time,
+                error=str(e),
+            )
+        finally:
+            if not self.keep_artifacts:
+                try:
+                    shutil.rmtree(workspace)
+                except Exception:
+                    pass
+
+
+def run_live_interactive_tests(
+    scenarios: List[InteractiveScenario],
+    model: str = "",
+    artifacts_dir: Optional[Path] = None,
+    verbose: bool = True,
+) -> RunSummary:
+    """
+    Convenience function to run live interactive tests.
+    
+    Args:
+        scenarios: List of scenarios to run
+        model: LLM model (default: gpt-4o-mini)
+        artifacts_dir: Directory for artifacts
+        verbose: Print progress
+        
+    Returns:
+        RunSummary with all results
+    """
+    runner = LiveInteractiveRunner(
+        model=model,
+        artifacts_dir=artifacts_dir,
+        verbose=verbose,
+    )
+    return runner.run_all(scenarios)

src/praisonai/tests/live/interactive/scenarios.py

@@ -0,0 +1,328 @@
+"""Interactive mode test scenarios.
+
+20 scenarios covering all interactive mode capabilities:
+- Basic chat (3)
+- File operations (5)
+- Code intelligence (4)
+- Multi-step workflows (4)
+- Multi-agent (2)
+- Edge cases (2)
+"""
+
+from typing import List
+
+from .runner import InteractiveScenario
+
+
+# =============================================================================
+# BASIC CHAT SCENARIOS (3)
+# =============================================================================
+
+SCENARIO_01_HELLO = InteractiveScenario(
+    id="int_01",
+    name="Basic Hello",
+    description="Test basic chat response without tools",
+    prompts=["Hello! What is 2 + 2?"],
+    expected_tools=[],
+    expected_response=r"4",
+    timeout=30,
+)
+
+SCENARIO_02_REASONING = InteractiveScenario(
+    id="int_02",
+    name="Simple Reasoning",
+    description="Test reasoning capability",
+    prompts=["If I have 3 apples and give away 1, how many do I have left?"],
+    expected_tools=[],
+    expected_response=r"2",
+    timeout=30,
+)
+
+SCENARIO_03_EXPLAIN = InteractiveScenario(
+    id="int_03",
+    name="Code Explanation",
+    description="Test code explanation without file access",
+    prompts=["Explain what this Python code does: print('Hello, World!')"],
+    expected_tools=[],
+    expected_response=r"print|Hello|output",
+    timeout=30,
+)
+
+
+# =============================================================================
+# FILE OPERATIONS SCENARIOS (5)
+# =============================================================================
+
+SCENARIO_04_CREATE_FILE = InteractiveScenario(
+    id="int_04",
+    name="Create File",
+    description="Test creating a new file with content",
+    prompts=["Create a file called hello.py with the content: print('hello world')"],
+    expected_tools=["acp_create_file"],
+    expected_files={"hello.py": r"print.*hello"},
+    timeout=60,
+)
+
+SCENARIO_05_READ_FILE = InteractiveScenario(
+    id="int_05",
+    name="Read File",
+    description="Test reading an existing file",
+    prompts=["Read the file README.md and tell me what it says"],
+    expected_tools=["read_file"],
+    expected_response=r"Test Project|README",
+    workspace_fixture="seeded",
+    timeout=60,
+)
+
+SCENARIO_06_LIST_FILES = InteractiveScenario(
+    id="int_06",
+    name="List Files",
+    description="Test listing files in directory",
+    prompts=["List all files in the current directory"],
+    expected_tools=["list_files"],
+    expected_response=r"README|main\.py|config",
+    workspace_fixture="seeded",
+    timeout=60,
+)
+
+SCENARIO_07_EDIT_FILE = InteractiveScenario(
+    id="int_07",
+    name="Edit File",
+    description="Test editing an existing file",
+    prompts=[
+        "Create a file called counter.py with: count = 1",
+        "Edit counter.py to change count = 1 to count = 10",
+    ],
+    expected_tools=["acp_create_file", "acp_edit_file"],
+    expected_files={"counter.py": r"count\s*=\s*10"},
+    timeout=90,
+)
+
+SCENARIO_08_DELETE_FILE = InteractiveScenario(
+    id="int_08",
+    name="Delete File",
+    description="Test deleting a file",
+    prompts=[
+        "Create a file called temp.txt with content: temporary",
+        "Delete the file temp.txt",
+    ],
+    expected_tools=["acp_create_file", "acp_delete_file"],
+    expected_response=r"deleted|removed",
+    timeout=60,
+)
+
+
+# =============================================================================
+# CODE INTELLIGENCE SCENARIOS (4)
+# =============================================================================
+
+SCENARIO_09_LIST_SYMBOLS = InteractiveScenario(
+    id="int_09",
+    name="List Symbols",
+    description="Test listing symbols in a Python file",
+    prompts=[
+        "Create a file calc.py with:\ndef add(a, b):\n    return a + b\n\ndef subtract(a, b):\n    return a - b",
+        "List all functions in calc.py",
+    ],
+    expected_tools=["acp_create_file", "lsp_list_symbols"],
+    expected_response=r"add|subtract",
+    timeout=90,
+)
+
+SCENARIO_10_FIND_DEFINITION = InteractiveScenario(
+    id="int_10",
+    name="Find Definition",
+    description="Test finding symbol definition",
+    prompts=["Find where the function 'main' is defined in main.py"],
+    expected_tools=["lsp_find_definition"],
+    workspace_fixture="python_project",
+    timeout=60,
+)
+
+SCENARIO_11_FIND_REFERENCES = InteractiveScenario(
+    id="int_11",
+    name="Find References",
+    description="Test finding symbol references",
+    prompts=["Find all references to the 'add' function in utils.py"],
+    expected_tools=["lsp_find_references"],
+    workspace_fixture="python_project",
+    timeout=60,
+)
+
+SCENARIO_12_EXECUTE_COMMAND = InteractiveScenario(
+    id="int_12",
+    name="Execute Command",
+    description="Test executing a shell command",
+    prompts=["Run the command: echo 'Hello from shell'"],
+    expected_tools=[],  # May use acp_execute_command or execute_command
+    expected_response=r"Hello.*shell|echo",
+    timeout=30,
+)
+
+
+# =============================================================================
+# MULTI-STEP WORKFLOW SCENARIOS (4)
+# =============================================================================
+
+SCENARIO_13_CREATE_AND_READ = InteractiveScenario(
+    id="int_13",
+    name="Create and Read",
+    description="Test creating a file then reading it back",
+    prompts=[
+        "Create a file called greeting.txt with: Hello, PraisonAI!",
+        "Read the file greeting.txt and confirm its contents",
+    ],
+    expected_tools=["acp_create_file", "read_file"],
+    expected_files={"greeting.txt": r"Hello.*PraisonAI"},
+    expected_response=r"Hello.*PraisonAI",
+    timeout=90,
+)
+
+SCENARIO_14_REFACTOR_WORKFLOW = InteractiveScenario(
+    id="int_14",
+    name="Refactor Workflow",
+    description="Test a refactoring workflow: read, analyze, edit",
+    prompts=[
+        "Read main.py and tell me what it does",
+        "Add a docstring to the main function in main.py",
+    ],
+    expected_tools=["read_file"],  # Edit tool varies
+    expected_files={"main.py": r'""".*"""|docstring|Main'},
+    workspace_fixture="python_project",
+    timeout=120,
+)
+
+SCENARIO_15_COPY_FILE = InteractiveScenario(
+    id="int_15",
+    name="Copy File Content",
+    description="Test reading one file and creating another with same content",
+    prompts=[
+        "Read the file main.py and show me its content",
+        "Now create a new file called backup.py with a simple print statement",
+    ],
+    expected_tools=["read_file"],  # Create tool varies
+    expected_files={"backup.py": r"print"},
+    workspace_fixture="python_project",
+    timeout=90,
+)
+
+SCENARIO_16_ANALYZE_AND_FIX = InteractiveScenario(
+    id="int_16",
+    name="Analyze and Fix",
+    description="Test analyzing code and making improvements",
+    prompts=[
+        "Create a file buggy.py with:\ndef divide(a, b):\n    return a / b",
+        "Fix buggy.py to handle division by zero by raising ValueError",
+    ],
+    expected_tools=["acp_create_file", "acp_edit_file"],
+    expected_files={"buggy.py": r"ValueError|zero|if.*b.*==.*0"},
+    timeout=90,
+)
+
+
+# =============================================================================
+# MULTI-AGENT SCENARIOS (2)
+# =============================================================================
+
+SCENARIO_17_SIMPLE_HANDOFF = InteractiveScenario(
+    id="int_17",
+    name="Simple Task Delegation",
+    description="Test handling a task that requires multiple steps",
+    prompts=[
+        "First, create a file called task.py with: # TODO: implement",
+        "Then, edit task.py to add a function called process() that returns 'done'",
+    ],
+    expected_tools=["acp_create_file", "acp_edit_file"],
+    expected_files={"task.py": r"def process.*return.*done"},
+    timeout=120,
+)
+
+SCENARIO_18_RESEARCH_AND_CODE = InteractiveScenario(
+    id="int_18",
+    name="Research and Code",
+    description="Test combining information gathering with code generation",
+    prompts=[
+        "List the files in the current directory",
+        "Based on what you see, create a new file called summary.txt describing the project structure",
+    ],
+    expected_tools=["list_files", "acp_create_file"],
+    expected_files={"summary.txt": r"file|directory|project"},
+    workspace_fixture="seeded",
+    timeout=90,
+)
+
+
+# =============================================================================
+# EDGE CASE SCENARIOS (2)
+# =============================================================================
+
+SCENARIO_19_EMPTY_RESPONSE = InteractiveScenario(
+    id="int_19",
+    name="Handle Empty Input",
+    description="Test handling minimal input gracefully",
+    prompts=["Hi"],
+    expected_tools=[],
+    timeout=30,
+)
+
+SCENARIO_20_COMPLEX_PROMPT = InteractiveScenario(
+    id="int_20",
+    name="Complex Multi-Part Request",
+    description="Test handling a complex request with multiple parts",
+    prompts=[
+        "Create a Python module called mathops.py with two functions: add(a, b) that returns a+b, and multiply(a, b) that returns a*b. Include docstrings for each function.",
+    ],
+    expected_tools=["acp_create_file"],
+    expected_files={"mathops.py": r"def add.*def multiply|def multiply.*def add"},
+    timeout=90,
+)
+
+
+# =============================================================================
+# ALL SCENARIOS
+# =============================================================================
+
+ALL_SCENARIOS: List[InteractiveScenario] = [
+    # Basic Chat (3)
+    SCENARIO_01_HELLO,
+    SCENARIO_02_REASONING,
+    SCENARIO_03_EXPLAIN,
+    # File Operations (5)
+    SCENARIO_04_CREATE_FILE,
+    SCENARIO_05_READ_FILE,
+    SCENARIO_06_LIST_FILES,
+    SCENARIO_07_EDIT_FILE,
+    SCENARIO_08_DELETE_FILE,
+    # Code Intelligence (4)
+    SCENARIO_09_LIST_SYMBOLS,
+    SCENARIO_10_FIND_DEFINITION,
+    SCENARIO_11_FIND_REFERENCES,
+    SCENARIO_12_EXECUTE_COMMAND,
+    # Multi-step Workflows (4)
+    SCENARIO_13_CREATE_AND_READ,
+    SCENARIO_14_REFACTOR_WORKFLOW,
+    SCENARIO_15_COPY_FILE,
+    SCENARIO_16_ANALYZE_AND_FIX,
+    # Multi-agent (2)
+    SCENARIO_17_SIMPLE_HANDOFF,
+    SCENARIO_18_RESEARCH_AND_CODE,
+    # Edge Cases (2)
+    SCENARIO_19_EMPTY_RESPONSE,
+    SCENARIO_20_COMPLEX_PROMPT,
+]
+
+
+def get_scenarios_by_category(category: str) -> List[InteractiveScenario]:
+    """Get scenarios by category."""
+    categories = {
+        "basic": [SCENARIO_01_HELLO, SCENARIO_02_REASONING, SCENARIO_03_EXPLAIN],
+        "files": [SCENARIO_04_CREATE_FILE, SCENARIO_05_READ_FILE, SCENARIO_06_LIST_FILES,
+                  SCENARIO_07_EDIT_FILE, SCENARIO_08_DELETE_FILE],
+        "code": [SCENARIO_09_LIST_SYMBOLS, SCENARIO_10_FIND_DEFINITION,
+                 SCENARIO_11_FIND_REFERENCES, SCENARIO_12_EXECUTE_COMMAND],
+        "workflow": [SCENARIO_13_CREATE_AND_READ, SCENARIO_14_REFACTOR_WORKFLOW,
+                     SCENARIO_15_COPY_FILE, SCENARIO_16_ANALYZE_AND_FIX],
+        "multi": [SCENARIO_17_SIMPLE_HANDOFF, SCENARIO_18_RESEARCH_AND_CODE],
+        "edge": [SCENARIO_19_EMPTY_RESPONSE, SCENARIO_20_COMPLEX_PROMPT],
+    }
+    return categories.get(category, [])

src/praisonai/tests/live/interactive/test_smoke.py

@@ -0,0 +1,130 @@
+"""Live smoke tests for interactive mode.
+
+These tests verify PraisonAI's interactive mode using real API calls.
+Tests use HeadlessInteractiveCore to execute prompts through the same
+pipeline as the interactive TUI.
+
+Run with: PRAISONAI_LIVE_INTERACTIVE=1 pytest tests/live/interactive/test_smoke.py -v
+"""
+
+import pytest
+
+from .runner import (
+    is_live_interactive_enabled,
+    has_api_key,
+    LiveInteractiveRunner,
+)
+from .scenarios import ALL_SCENARIOS, get_scenarios_by_category
+
+
+pytestmark = pytest.mark.skipif(
+    not is_live_interactive_enabled() or not has_api_key(),
+    reason="Live interactive tests disabled (set PRAISONAI_LIVE_INTERACTIVE=1 and OPENAI_API_KEY)"
+)
+
+
+@pytest.fixture(scope="module")
+def runner():
+    """Create a shared runner for all tests."""
+    return LiveInteractiveRunner(verbose=True)
+
+
+class TestLiveInteractiveSmoke:
+    """Live smoke tests for interactive mode scenarios."""
+
+    @pytest.mark.parametrize("scenario", ALL_SCENARIOS, ids=lambda s: s.id)
+    def test_scenario(self, scenario, runner):
+        """Run a single interactive scenario with retry."""
+        result = runner._run_with_retry(scenario)
+        
+        if not result.passed:
+            print(f"\n{'='*60}")
+            print(f"SCENARIO FAILED: {scenario.id} - {scenario.name}")
+            print(f"{'='*60}")
+            print(f"Description: {scenario.description}")
+            print(f"Prompts: {scenario.prompts}")
+            print(f"Attempts: {result.attempts}")
+            print(f"Assertions: {result.assertions}")
+            print(f"Tool calls: {result.tool_calls}")
+            print(f"Error: {result.error}")
+            print(f"\nResponse (last 500 chars):\n{result.response[-500:] if result.response else 'None'}")
+            print(f"{'='*60}\n")
+        
+        assert result.passed, f"Scenario {scenario.id} failed after {result.attempts} attempts: {result.error}"
+
+
+class TestBasicChat:
+    """Basic chat scenarios."""
+
+    @pytest.mark.parametrize("scenario", get_scenarios_by_category("basic"), ids=lambda s: s.id)
+    def test_basic(self, scenario, runner):
+        """Test basic chat scenarios."""
+        result = runner._run_with_retry(scenario)
+        assert result.passed, f"Failed: {result.error}"
+
+
+class TestFileOperations:
+    """File operation scenarios."""
+
+    @pytest.mark.parametrize("scenario", get_scenarios_by_category("files"), ids=lambda s: s.id)
+    def test_files(self, scenario, runner):
+        """Test file operation scenarios."""
+        result = runner._run_with_retry(scenario)
+        assert result.passed, f"Failed: {result.error}"
+
+
+class TestCodeIntelligence:
+    """Code intelligence scenarios."""
+
+    @pytest.mark.parametrize("scenario", get_scenarios_by_category("code"), ids=lambda s: s.id)
+    def test_code(self, scenario, runner):
+        """Test code intelligence scenarios."""
+        result = runner._run_with_retry(scenario)
+        assert result.passed, f"Failed: {result.error}"
+
+
+class TestWorkflows:
+    """Multi-step workflow scenarios."""
+
+    @pytest.mark.parametrize("scenario", get_scenarios_by_category("workflow"), ids=lambda s: s.id)
+    def test_workflow(self, scenario, runner):
+        """Test workflow scenarios."""
+        result = runner._run_with_retry(scenario)
+        assert result.passed, f"Failed: {result.error}"
+
+
+class TestMultiAgent:
+    """Multi-agent scenarios."""
+
+    @pytest.mark.parametrize("scenario", get_scenarios_by_category("multi"), ids=lambda s: s.id)
+    def test_multi(self, scenario, runner):
+        """Test multi-agent scenarios."""
+        result = runner._run_with_retry(scenario)
+        assert result.passed, f"Failed: {result.error}"
+
+
+class TestEdgeCases:
+    """Edge case scenarios."""
+
+    @pytest.mark.parametrize("scenario", get_scenarios_by_category("edge"), ids=lambda s: s.id)
+    def test_edge(self, scenario, runner):
+        """Test edge case scenarios."""
+        result = runner._run_with_retry(scenario)
+        assert result.passed, f"Failed: {result.error}"
+
+
+def run_all_and_report():
+    """Run all scenarios and print report (for direct execution)."""
+    from .scenarios import ALL_SCENARIOS
+    
+    runner = LiveInteractiveRunner(verbose=True)
+    summary = runner.run_all(ALL_SCENARIOS)
+    summary.print_summary()
+    
+    return summary.pass_rate == 1.0
+
+
+if __name__ == "__main__":
+    import sys
+    success = run_all_and_report()
+    sys.exit(0 if success else 1)

src/praisonai/tests/unit/cli/test_approval_modes.py

@@ -0,0 +1,273 @@
+"""
+Tests for Approval Mode Matrix.
+
+Tests that approval modes (auto, manual, scoped) work correctly
+with the interactive runtime and tool execution.
+"""
+
+import os
+import tempfile
+
+
+class TestRuntimeConfigApprovalMode:
+    """Tests for RuntimeConfig approval_mode setting."""
+    
+    def test_default_approval_mode(self):
+        """Test default approval mode is auto."""
+        from praisonai.cli.features.interactive_runtime import RuntimeConfig
+        
+        config = RuntimeConfig()
+        assert config.approval_mode == "auto"
+    
+    def test_custom_approval_mode(self):
+        """Test setting custom approval mode."""
+        from praisonai.cli.features.interactive_runtime import RuntimeConfig
+        
+        config = RuntimeConfig(approval_mode="manual")
+        assert config.approval_mode == "manual"
+        
+        config = RuntimeConfig(approval_mode="scoped")
+        assert config.approval_mode == "scoped"
+
+
+class TestInteractiveRuntimeReadOnly:
+    """Tests for InteractiveRuntime read_only property."""
+    
+    def test_auto_mode_not_read_only(self):
+        """Test that auto mode is not read-only."""
+        from praisonai.cli.features.interactive_runtime import (
+            InteractiveRuntime,
+            RuntimeConfig,
+        )
+        
+        config = RuntimeConfig(
+            workspace="/tmp/test",
+            approval_mode="auto",
+        )
+        runtime = InteractiveRuntime(config)
+        
+        # Auto mode should NOT be read-only
+        assert runtime.read_only is False
+    
+    def test_auto_mode_writable_without_acp(self):
+        """Test that auto mode allows writes even without ACP."""
+        from praisonai.cli.features.interactive_runtime import (
+            InteractiveRuntime,
+            RuntimeConfig,
+        )
+        
+        config = RuntimeConfig(
+            workspace="/tmp/test",
+            approval_mode="auto",
+            acp_enabled=False,  # ACP disabled
+        )
+        runtime = InteractiveRuntime(config)
+        runtime._read_only = True  # Simulate ACP not ready
+        
+        # Auto mode should still NOT be read-only
+        assert runtime.read_only is False
+    
+    def test_manual_mode_read_only_without_acp(self):
+        """Test that manual mode is read-only without ACP."""
+        from praisonai.cli.features.interactive_runtime import (
+            InteractiveRuntime,
+            RuntimeConfig,
+        )
+        
+        config = RuntimeConfig(
+            workspace="/tmp/test",
+            approval_mode="manual",
+        )
+        runtime = InteractiveRuntime(config)
+        runtime._read_only = True  # Simulate ACP not ready
+        
+        # Manual mode should be read-only when ACP not ready
+        assert runtime.read_only is True
+    
+    def test_scoped_mode_read_only_without_acp(self):
+        """Test that scoped mode is read-only without ACP."""
+        from praisonai.cli.features.interactive_runtime import (
+            InteractiveRuntime,
+            RuntimeConfig,
+        )
+        
+        config = RuntimeConfig(
+            workspace="/tmp/test",
+            approval_mode="scoped",
+        )
+        runtime = InteractiveRuntime(config)
+        runtime._read_only = True  # Simulate ACP not ready
+        
+        # Scoped mode should be read-only when ACP not ready
+        assert runtime.read_only is True
+    
+    def test_env_var_auto_mode_override(self):
+        """Test that PRAISON_APPROVAL_MODE env var can override to auto mode."""
+        from praisonai.cli.features.interactive_runtime import (
+            InteractiveRuntime,
+            RuntimeConfig,
+        )
+        
+        # Save and clear env var first
+        original = os.environ.get("PRAISON_APPROVAL_MODE")
+        if "PRAISON_APPROVAL_MODE" in os.environ:
+            del os.environ["PRAISON_APPROVAL_MODE"]
+        
+        try:
+            # Without env var, manual mode should be read-only when _read_only=True
+            config = RuntimeConfig(
+                workspace="/tmp/test",
+                approval_mode="manual",
+            )
+            runtime = InteractiveRuntime(config)
+            runtime._read_only = True
+            
+            # Manual mode without env override should be read-only
+            assert runtime.read_only is True
+            
+            # Now set env var to auto - this should override
+            os.environ["PRAISON_APPROVAL_MODE"] = "auto"
+            
+            # Create new runtime with manual config but env=auto
+            config2 = RuntimeConfig(
+                workspace="/tmp/test",
+                approval_mode="manual",
+            )
+            runtime2 = InteractiveRuntime(config2)
+            runtime2._read_only = True
+            
+            # The read_only property checks config first, then env
+            # Config is manual, so it checks env var next
+            # Env var is auto, so it returns False (not read-only)
+            assert runtime2.read_only is False
+            
+        finally:
+            if original:
+                os.environ["PRAISON_APPROVAL_MODE"] = original
+            elif "PRAISON_APPROVAL_MODE" in os.environ:
+                del os.environ["PRAISON_APPROVAL_MODE"]
+
+
+class TestToolConfigApprovalMode:
+    """Tests for ToolConfig approval_mode setting."""
+    
+    def test_default_approval_mode(self):
+        """Test default approval mode in ToolConfig."""
+        from praisonai.cli.features.interactive_tools import ToolConfig
+        
+        config = ToolConfig()
+        assert config.approval_mode == "auto"
+    
+    def test_from_env_approval_mode(self):
+        """Test ToolConfig.from_env reads approval mode."""
+        from praisonai.cli.features.interactive_tools import ToolConfig
+        
+        original = os.environ.get("PRAISON_APPROVAL_MODE")
+        os.environ["PRAISON_APPROVAL_MODE"] = "manual"
+        
+        try:
+            config = ToolConfig.from_env()
+            assert config.approval_mode == "manual"
+        finally:
+            if original:
+                os.environ["PRAISON_APPROVAL_MODE"] = original
+            elif "PRAISON_APPROVAL_MODE" in os.environ:
+                del os.environ["PRAISON_APPROVAL_MODE"]
+
+
+class TestHeadlessConfigApprovalMode:
+    """Tests for HeadlessConfig approval_mode setting."""
+    
+    def test_default_approval_mode(self):
+        """Test default approval mode in HeadlessConfig."""
+        from praisonai.cli.features.interactive_core_headless import HeadlessConfig
+        
+        config = HeadlessConfig()
+        assert config.approval_mode == "auto"
+    
+    def test_custom_approval_mode(self):
+        """Test custom approval mode in HeadlessConfig."""
+        from praisonai.cli.features.interactive_core_headless import HeadlessConfig
+        
+        config = HeadlessConfig(approval_mode="manual")
+        assert config.approval_mode == "manual"
+
+
+class TestApprovalModeIntegration:
+    """Integration tests for approval mode across components."""
+    
+    def test_auto_mode_allows_file_creation_via_tools(self):
+        """Test that auto mode allows file creation through tools."""
+        from praisonai.cli.features.interactive_tools import (
+            get_interactive_tools,
+            ToolConfig,
+        )
+        
+        with tempfile.TemporaryDirectory() as tmpdir:
+            config = ToolConfig(
+                workspace=tmpdir,
+                approval_mode="auto",
+                enable_acp=True,
+                enable_lsp=False,
+                enable_basic=True,
+            )
+            
+            # Get tools - this should work without error
+            tools = get_interactive_tools(config=config)
+            
+            # Should have ACP tools
+            tool_names = [t.__name__ for t in tools]
+            assert "acp_create_file" in tool_names or len(tools) > 0
+    
+    def test_tool_config_propagates_to_runtime(self):
+        """Test that ToolConfig approval_mode propagates to runtime."""
+        from praisonai.cli.features.interactive_tools import ToolConfig
+        from praisonai.cli.features.interactive_runtime import RuntimeConfig
+        
+        tool_config = ToolConfig(approval_mode="scoped")
+        
+        # When creating RuntimeConfig from ToolConfig values
+        runtime_config = RuntimeConfig(
+            workspace=tool_config.workspace,
+            approval_mode=tool_config.approval_mode,
+        )
+        
+        assert runtime_config.approval_mode == "scoped"
+
+
+class TestApprovalModeMatrix:
+    """Matrix tests for all approval mode combinations."""
+    
+    def test_approval_mode_matrix(self):
+        """Test all approval mode combinations."""
+        from praisonai.cli.features.interactive_runtime import (
+            InteractiveRuntime,
+            RuntimeConfig,
+        )
+        
+        # Matrix: (approval_mode, acp_ready, expected_read_only)
+        test_cases = [
+            # Auto mode is never read-only
+            ("auto", True, False),
+            ("auto", False, False),
+            # Manual mode depends on _read_only flag
+            ("manual", True, False),  # ACP ready, not read-only
+            ("manual", False, True),  # ACP not ready, read-only
+            # Scoped mode depends on _read_only flag
+            ("scoped", True, False),
+            ("scoped", False, True),
+        ]
+        
+        for approval_mode, acp_ready, expected_read_only in test_cases:
+            config = RuntimeConfig(
+                workspace="/tmp/test",
+                approval_mode=approval_mode,
+            )
+            runtime = InteractiveRuntime(config)
+            runtime._read_only = not acp_ready
+            
+            actual = runtime.read_only
+            assert actual == expected_read_only, (
+                f"Failed for approval_mode={approval_mode}, acp_ready={acp_ready}: "
+                f"expected read_only={expected_read_only}, got {actual}"
+            )

src/praisonai/tests/unit/cli/test_async_tui.py

@@ -0,0 +1,848 @@
+"""
+Tests for PraisonAI Async TUI Application.
+"""
+
+
+class TestAsyncTUIConfig:
+    """Tests for AsyncTUIConfig dataclass."""
+    
+    def test_default_values(self):
+        """Test default configuration values."""
+        from praisonai.cli.interactive.async_tui import AsyncTUIConfig
+        
+        config = AsyncTUIConfig()
+        assert config.model == "gpt-4o-mini"
+        assert config.show_logo is True
+        assert config.show_status_bar is True
+        assert config.session_id is None
+        assert config.workspace is None
+    
+    def test_custom_values(self):
+        """Test custom configuration values."""
+        from praisonai.cli.interactive.async_tui import AsyncTUIConfig
+        
+        config = AsyncTUIConfig(
+            model="gpt-4o",
+            show_logo=False,
+            session_id="test123",
+        )
+        assert config.model == "gpt-4o"
+        assert config.show_logo is False
+        assert config.session_id == "test123"
+
+
+class TestChatMessage:
+    """Tests for ChatMessage dataclass."""
+    
+    def test_message_creation(self):
+        """Test message creation."""
+        from praisonai.cli.interactive.async_tui import ChatMessage
+        
+        msg = ChatMessage(role="user", content="Hello")
+        assert msg.role == "user"
+        assert msg.content == "Hello"
+        assert msg.timestamp is not None
+    
+    def test_message_roles(self):
+        """Test different message roles."""
+        from praisonai.cli.interactive.async_tui import ChatMessage
+        
+        roles = ["user", "assistant", "system", "status"]
+        for role in roles:
+            msg = ChatMessage(role=role, content="test")
+            assert msg.role == role
+
+
+class TestGetLogo:
+    """Tests for get_logo function."""
+    
+    def test_logo_wide_terminal(self):
+        """Test logo for wide terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_LARGE
+        
+        logo = get_logo(100)
+        assert logo == LOGO_LARGE
+    
+    def test_logo_medium_terminal(self):
+        """Test logo for medium terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_MEDIUM
+        
+        logo = get_logo(50)
+        assert logo == LOGO_MEDIUM
+    
+    def test_logo_narrow_terminal(self):
+        """Test logo for narrow terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_SMALL
+        
+        logo = get_logo(30)
+        assert logo == LOGO_SMALL
+    
+    def test_logo_contains_praison_ai(self):
+        """Test that logo contains 'Praison AI' branding."""
+        from praisonai.cli.branding import LOGO_SMALL
+        
+        assert "Praison AI" in LOGO_SMALL
+
+
+class TestAsyncTUI:
+    """Tests for AsyncTUI class."""
+    
+    def test_init_default(self):
+        """Test default initialization."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert tui.config is not None
+        assert tui.messages == []
+        assert tui._running is False
+        assert tui._processing is False
+        assert tui.session_id is not None
+    
+    def test_init_with_config(self):
+        """Test initialization with config."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, AsyncTUIConfig
+        
+        config = AsyncTUIConfig(model="gpt-4o", show_logo=False)
+        tui = AsyncTUI(config=config)
+        assert tui.config.model == "gpt-4o"
+        assert tui.config.show_logo is False
+    
+    def test_format_output_empty(self):
+        """Test format_output with no messages."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, AsyncTUIConfig
+        
+        config = AsyncTUIConfig(show_logo=True)
+        tui = AsyncTUI(config=config)
+        
+        output = tui._format_output()
+        assert "Type your message" in output or "gpt-4o-mini" in output
+    
+    def test_format_output_with_messages(self):
+        """Test format_output with messages."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, ChatMessage
+        
+        tui = AsyncTUI()
+        tui.messages = [
+            ChatMessage(role="user", content="Hello"),
+            ChatMessage(role="assistant", content="Hi there"),
+        ]
+        
+        output = tui._format_output()
+        assert "Hello" in output
+        assert "Hi there" in output
+    
+    def test_format_output_with_processing(self):
+        """Test format_output shows processing indicator."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._processing = True
+        tui._status_text = "Thinking..."
+        
+        output = tui._format_output()
+        assert "Thinking..." in output
+    
+    def test_handle_command_exit(self):
+        """Test handling exit command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._running = True
+        
+        result = tui._handle_command("/exit")
+        assert result is True
+        assert tui._running is False
+    
+    def test_handle_command_quit(self):
+        """Test handling quit command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._running = True
+        
+        result = tui._handle_command("/quit")
+        assert result is True
+        assert tui._running is False
+    
+    def test_handle_command_help(self):
+        """Test handling help command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/help")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert tui.messages[0].role == "system"
+        assert "PageUp" in tui.messages[0].content  # Scroll instructions
+    
+    def test_handle_command_clear(self):
+        """Test handling clear command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, ChatMessage
+        
+        tui = AsyncTUI()
+        tui.messages = [ChatMessage(role="user", content="test")]
+        
+        result = tui._handle_command("/clear")
+        assert result is True
+        # Clear now adds a confirmation message
+        assert len(tui.messages) == 1
+        assert "cleared" in tui.messages[0].content.lower()
+    
+    def test_handle_command_new(self):
+        """Test handling new command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, ChatMessage
+        
+        tui = AsyncTUI()
+        old_session = tui.session_id
+        tui.messages = [ChatMessage(role="user", content="test")]
+        
+        result = tui._handle_command("/new")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert tui.session_id != old_session
+    
+    def test_handle_command_model_show(self):
+        """Test handling model command (show)."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/model")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "gpt-4o-mini" in tui.messages[0].content
+    
+    def test_handle_command_model_change(self):
+        """Test handling model command (change)."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/model gpt-4o")
+        assert result is True
+        assert tui.config.model == "gpt-4o"
+        assert tui._agent is None
+    
+    def test_handle_command_unknown(self):
+        """Test handling unknown command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/unknown")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "Unknown" in tui.messages[0].content
+
+
+class TestStartAsyncTUI:
+    """Tests for start_async_tui function."""
+    
+    def test_function_exists(self):
+        """Test that start_async_tui function exists."""
+        from praisonai.cli.interactive.async_tui import start_async_tui
+        assert callable(start_async_tui)
+
+
+class TestASCIILogos:
+    """Tests for ASCII logo constants."""
+    
+    def test_logo_exists(self):
+        """Test LOGO constant exists and has content."""
+        from praisonai.cli.branding import LOGO_LARGE
+        assert len(LOGO_LARGE) > 0
+        assert "██" in LOGO_LARGE
+    
+    def test_logo_small_exists(self):
+        """Test LOGO_SMALL constant exists."""
+        from praisonai.cli.branding import LOGO_SMALL
+        assert len(LOGO_SMALL) > 0
+    
+    def test_logo_minimal_branding(self):
+        """Test LOGO_MINIMAL has correct branding."""
+        from praisonai.cli.branding import LOGO_MINIMAL
+        assert "Praison AI" in LOGO_MINIMAL
+
+
+class TestMessageFormatting:
+    """Tests for message formatting."""
+    
+    def test_user_message_prefix(self):
+        """Test user messages have correct prefix."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, ChatMessage
+        
+        tui = AsyncTUI()
+        tui.messages = [ChatMessage(role="user", content="test")]
+        
+        output = tui._format_output()
+        assert "› test" in output
+    
+    def test_assistant_message_prefix(self):
+        """Test assistant messages have correct prefix."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, ChatMessage
+        
+        tui = AsyncTUI()
+        tui.messages = [ChatMessage(role="assistant", content="response")]
+        
+        output = tui._format_output()
+        assert "● response" in output
+    
+    def test_status_message_prefix(self):
+        """Test status messages have correct prefix."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, ChatMessage
+        
+        tui = AsyncTUI()
+        tui.messages = [ChatMessage(role="status", content="thinking")]
+        
+        output = tui._format_output()
+        assert "⏳ thinking" in output
+
+
+class TestNonBlockingExecution:
+    """Tests for non-blocking execution features."""
+    
+    def test_processing_flag(self):
+        """Test processing flag is used correctly."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert tui._processing is False
+        
+        tui._processing = True
+        tui._status_text = "Working..."
+        
+        output = tui._format_output()
+        assert "Working..." in output
+    
+    def test_status_text_updates(self):
+        """Test status text can be updated."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._processing = True
+        
+        tui._status_text = "Step 1"
+        output1 = tui._format_output()
+        assert "Step 1" in output1
+        
+        tui._status_text = "Step 2"
+        output2 = tui._format_output()
+        assert "Step 2" in output2
+
+
+class TestNewSlashCommands:
+    """Tests for newly added slash commands."""
+    
+    def test_session_command(self):
+        """Test /session command shows session info."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/session")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "Session:" in tui.messages[0].content
+        assert tui.session_id in tui.messages[0].content
+    
+    def test_history_command_empty(self):
+        """Test /history command with no history."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/history")
+        assert result is True
+        assert "No conversation history" in tui.messages[0].content
+    
+    def test_history_command_with_history(self):
+        """Test /history command with conversation history."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._conversation_history = [
+            {"role": "user", "content": "Hello"},
+            {"role": "assistant", "content": "Hi there"},
+        ]
+        result = tui._handle_command("/history")
+        assert result is True
+        assert "[user]" in tui.messages[0].content
+    
+    def test_cost_command(self):
+        """Test /cost command shows token usage."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._total_tokens = 100
+        tui._total_cost = 0.001
+        result = tui._handle_command("/cost")
+        assert result is True
+        assert "100" in tui.messages[0].content
+    
+    def test_compact_command(self):
+        """Test /compact command toggles compact mode."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert tui.config.compact_mode is False
+        
+        result = tui._handle_command("/compact")
+        assert result is True
+        assert tui.config.compact_mode is True
+        
+        result = tui._handle_command("/compact")
+        assert tui.config.compact_mode is False
+    
+    def test_multiline_command(self):
+        """Test /multiline command toggles multiline mode."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert tui.config.multiline_mode is False
+        
+        result = tui._handle_command("/multiline")
+        assert result is True
+        assert tui.config.multiline_mode is True
+    
+    def test_files_command(self):
+        """Test /files command lists workspace files."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._workspace_files = ["test.py", "main.py"]
+        result = tui._handle_command("/files")
+        assert result is True
+        assert "@test.py" in tui.messages[0].content
+    
+    def test_queue_command_empty(self):
+        """Test /queue command with empty queue."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/queue")
+        assert result is True
+        assert "No prompts in queue" in tui.messages[0].content
+    
+    def test_queue_command_with_items(self):
+        """Test /queue command with queued prompts."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._prompt_queue = ["First prompt", "Second prompt"]
+        result = tui._handle_command("/queue")
+        assert result is True
+        assert "First prompt" in tui.messages[0].content
+
+
+class TestQueueSystem:
+    """Tests for prompt queue system."""
+    
+    def test_queue_initialization(self):
+        """Test queue is initialized empty."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert tui._prompt_queue == []
+    
+    def test_queue_or_execute_when_not_processing(self):
+        """Test _queue_or_execute executes immediately when not processing."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tui._processing = False
+        
+        # This would normally start background execution
+        # Just test that it adds user message
+        tui._queue_or_execute("test prompt")
+        
+        assert len(tui.messages) == 1
+        assert tui.messages[0].role == "user"
+        assert tui.messages[0].content == "test prompt"
+
+
+class TestFileProcessing:
+    """Tests for @file mention processing."""
+    
+    def test_process_file_mentions_no_mentions(self):
+        """Test processing prompt with no @file mentions."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._process_file_mentions("Hello world")
+        assert result == "Hello world"
+    
+    def test_process_file_mentions_with_mention(self):
+        """Test processing prompt with @file mention."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        import tempfile
+        import os
+        
+        tui = AsyncTUI()
+        
+        # Create a temp file
+        with tempfile.NamedTemporaryFile(mode='w', suffix='.txt', delete=False) as f:
+            f.write("test content")
+            temp_path = f.name
+        
+        try:
+            tui.config.workspace = os.path.dirname(temp_path)
+            filename = os.path.basename(temp_path)
+            result = tui._process_file_mentions(f"Check @{filename}")
+            assert "test content" in result
+        finally:
+            os.unlink(temp_path)
+
+
+class TestWorkspaceScanning:
+    """Tests for workspace file scanning."""
+    
+    def test_workspace_files_initialized(self):
+        """Test workspace files list is initialized."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert isinstance(tui._workspace_files, list)
+    
+    def test_scan_workspace_limits_files(self):
+        """Test workspace scanning limits files (may exceed slightly due to batch processing)."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        # Verify the limit is approximately enforced (may exceed slightly due to batch)
+        assert len(tui._workspace_files) <= 1100  # Allow some margin
+
+
+class TestToolLoading:
+    """Tests for tool loading functionality."""
+    
+    def test_load_tools_returns_list(self):
+        """Test _load_tools returns a list."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        tools = tui._load_tools()
+        assert isinstance(tools, list)
+    
+    def test_load_tools_method_exists(self):
+        """Test _load_tools method exists."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert hasattr(tui, '_load_tools')
+        assert callable(tui._load_tools)
+
+
+class TestNewSessionCommands:
+    """Tests for session-related commands."""
+    
+    def test_sessions_command(self):
+        """Test /sessions command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/sessions")
+        assert result is True
+        # Should have a message about sessions
+        assert len(tui.messages) >= 1
+    
+    def test_continue_command(self):
+        """Test /continue command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/continue")
+        assert result is True
+        assert len(tui.messages) >= 1
+    
+    def test_import_command_no_args(self):
+        """Test /import command without arguments."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/import")
+        assert result is True
+        assert "Usage" in tui.messages[0].content
+    
+    def test_import_command_file_not_found(self):
+        """Test /import command with non-existent file."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/import nonexistent_file.json")
+        assert result is True
+        assert "not found" in tui.messages[0].content.lower() or "failed" in tui.messages[0].content.lower()
+    
+    def test_status_command(self):
+        """Test /status command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/status")
+        assert result is True
+        # Should show runtime status or "not initialized"
+        assert len(tui.messages) >= 1
+
+
+class TestRuntimeIntegration:
+    """Tests for runtime integration."""
+    
+    def test_runtime_fields_initialized(self):
+        """Test runtime fields are initialized."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert tui._runtime is None
+        assert tui._runtime_started is False
+    
+    def test_start_runtime_method_exists(self):
+        """Test _start_runtime method exists."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert hasattr(tui, '_start_runtime')
+        # It's an async method
+        import asyncio
+        assert asyncio.iscoroutinefunction(tui._start_runtime)
+
+
+class TestLogoDisplay:
+    """Tests for logo display functionality."""
+    
+    def test_logo_always_shows_when_enabled(self):
+        """Test logo shows even when messages exist."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, AsyncTUIConfig, ChatMessage
+        
+        config = AsyncTUIConfig(show_logo=True)
+        tui = AsyncTUI(config=config)
+        
+        # Add a message
+        tui.messages.append(ChatMessage(role="system", content="Test message"))
+        
+        # Logo should still be in output
+        output = tui._format_output()
+        assert "Praison" in output or "██" in output  # Logo contains these
+    
+    def test_logo_hidden_when_disabled(self):
+        """Test logo is hidden when show_logo=False."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, AsyncTUIConfig
+        
+        config = AsyncTUIConfig(show_logo=False)
+        tui = AsyncTUI(config=config)
+        
+        output = tui._format_output()
+        # Should not contain the large logo characters
+        assert "██████" not in output
+    
+    def test_tips_hidden_after_first_message(self):
+        """Test tips are hidden after first message but logo remains."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, AsyncTUIConfig, ChatMessage
+        
+        config = AsyncTUIConfig(show_logo=True)
+        tui = AsyncTUI(config=config)
+        
+        # Before messages - tips should show
+        output_before = tui._format_output()
+        assert "Type your message" in output_before
+        
+        # Add a message
+        tui.messages.append(ChatMessage(role="user", content="Hello"))
+        
+        # After messages - tips should be hidden but logo remains
+        output_after = tui._format_output()
+        assert "Type your message" not in output_after
+        # Logo should still be there
+        assert "Model:" in output_after
+
+
+class TestToolCallVisibility:
+    """Tests for tool call visibility."""
+    
+    def test_tool_role_in_format_output(self):
+        """Test tool role messages are formatted correctly."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, ChatMessage
+        
+        tui = AsyncTUI()
+        tui.messages.append(ChatMessage(role="tool", content="read_file completed"))
+        
+        output = tui._format_output()
+        assert "⚙" in output
+        assert "read_file completed" in output
+
+
+class TestDebugMode:
+    """Tests for debug mode (/debug command and --debug flag)."""
+    
+    def test_debug_command_toggles_mode(self):
+        """Test /debug command toggles debug mode."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        assert tui.config.debug is False
+        
+        result = tui._handle_command("/debug")
+        
+        assert result is True
+        assert tui.config.debug is True
+        assert "enabled" in tui.messages[0].content.lower()
+        assert "async_tui_debug.log" in tui.messages[0].content
+    
+    def test_debug_flag_in_config(self):
+        """Test debug flag can be set via config."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI, AsyncTUIConfig
+        
+        config = AsyncTUIConfig(debug=True)
+        tui = AsyncTUI(config=config)
+        
+        assert tui.config.debug is True
+
+
+class TestPlanningIntegration:
+    """Tests for planning integration (/plan command)."""
+    
+    def test_plan_command_no_args(self):
+        """Test /plan command without arguments shows usage."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/plan")
+        
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "Usage:" in tui.messages[0].content
+        assert "/plan" in tui.messages[0].content
+    
+    def test_plan_command_with_task(self):
+        """Test /plan command with a task description."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        # Mock _queue_or_execute to prevent actual execution
+        executed_prompts = []
+        tui._queue_or_execute = lambda p: executed_prompts.append(p)
+        
+        result = tui._handle_command("/plan refactor the auth module")
+        
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "Creating plan" in tui.messages[0].content
+        assert len(executed_prompts) == 1
+        assert "refactor the auth module" in executed_prompts[0]
+
+
+class TestHandoffSubagents:
+    """Tests for handoff/subagents (/handoff command)."""
+    
+    def test_handoff_command_no_args(self):
+        """Test /handoff command without arguments shows usage."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/handoff")
+        
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "Usage:" in tui.messages[0].content
+        assert "code" in tui.messages[0].content
+        assert "research" in tui.messages[0].content
+    
+    def test_handoff_command_missing_task(self):
+        """Test /handoff command with only agent type."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command("/handoff code")
+        
+        assert result is True
+        assert "specify both" in tui.messages[0].content.lower()
+    
+    def test_handoff_command_code_agent(self):
+        """Test /handoff code command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        executed_prompts = []
+        tui._queue_or_execute = lambda p: executed_prompts.append(p)
+        
+        result = tui._handle_command('/handoff code "fix the bug"')
+        
+        assert result is True
+        assert "Handing off" in tui.messages[0].content
+        assert len(executed_prompts) == 1
+        assert "code" in executed_prompts[0].lower()
+    
+    def test_handoff_command_research_agent(self):
+        """Test /handoff research command."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        executed_prompts = []
+        tui._queue_or_execute = lambda p: executed_prompts.append(p)
+        
+        result = tui._handle_command('/handoff research "find best practices"')
+        
+        assert result is True
+        assert len(executed_prompts) == 1
+        assert "research" in executed_prompts[0].lower()
+    
+    def test_handoff_command_unknown_agent(self):
+        """Test /handoff with unknown agent type."""
+        from praisonai.cli.interactive.async_tui import AsyncTUI
+        
+        tui = AsyncTUI()
+        result = tui._handle_command('/handoff unknown "some task"')
+        
+        assert result is True
+        assert "Unknown agent type" in tui.messages[0].content
+
+
+class TestInteractiveRuntimeReadOnly:
+    """Tests for InteractiveRuntime read_only property."""
+    
+    def test_read_only_false_when_config_approval_auto(self):
+        """Test read_only is False when config.approval_mode is 'auto'."""
+        from praisonai.cli.features.interactive_runtime import InteractiveRuntime, RuntimeConfig
+        
+        config = RuntimeConfig(approval_mode="auto")
+        runtime = InteractiveRuntime(config)
+        
+        # Even without ACP ready, should not be read-only in auto mode
+        assert runtime.read_only is False
+    
+    def test_read_only_false_when_env_approval_auto(self):
+        """Test read_only is False when PRAISON_APPROVAL_MODE env is 'auto'."""
+        import os
+        from praisonai.cli.features.interactive_runtime import InteractiveRuntime, RuntimeConfig
+        
+        # Set env var
+        old_val = os.environ.get("PRAISON_APPROVAL_MODE")
+        os.environ["PRAISON_APPROVAL_MODE"] = "auto"
+        
+        try:
+            config = RuntimeConfig(approval_mode="manual")  # Config says manual
+            runtime = InteractiveRuntime(config)
+            
+            # Config takes precedence, so this should be read-only
+            # But env var should also work as fallback
+            # Actually config.approval_mode="manual" so env var is checked
+            # Wait, config says manual, so it won't return False on first check
+            # Then env var check should return False
+            assert runtime.read_only is False
+        finally:
+            if old_val is None:
+                os.environ.pop("PRAISON_APPROVAL_MODE", None)
+            else:
+                os.environ["PRAISON_APPROVAL_MODE"] = old_val
+    
+    def test_read_only_true_when_manual_mode(self):
+        """Test read_only respects _read_only flag in manual mode."""
+        import os
+        from praisonai.cli.features.interactive_runtime import InteractiveRuntime, RuntimeConfig
+        
+        # Clear env var
+        old_val = os.environ.get("PRAISON_APPROVAL_MODE")
+        os.environ.pop("PRAISON_APPROVAL_MODE", None)
+        
+        try:
+            config = RuntimeConfig(approval_mode="manual")
+            runtime = InteractiveRuntime(config)
+            runtime._read_only = True  # Explicitly set read-only
+            
+            assert runtime.read_only is True
+        finally:
+            if old_val is not None:
+                os.environ["PRAISON_APPROVAL_MODE"] = old_val

src/praisonai/tests/unit/cli/test_branding.py

@@ -0,0 +1,117 @@
+"""
+Tests for the unified branding module.
+
+This module provides the single source of truth for logo/version/branding
+across all interactive UIs.
+"""
+
+import pytest
+
+
+class TestBrandingConstants:
+    """Tests for branding constants."""
+    
+    def test_product_name(self):
+        """Test PRODUCT_NAME constant."""
+        from praisonai.cli.branding import PRODUCT_NAME
+        assert PRODUCT_NAME == "Praison AI"
+    
+    def test_logo_large_exists(self):
+        """Test LOGO_LARGE constant exists and has ASCII art."""
+        from praisonai.cli.branding import LOGO_LARGE
+        assert len(LOGO_LARGE) > 0
+        assert "██" in LOGO_LARGE
+    
+    def test_logo_medium_exists(self):
+        """Test LOGO_MEDIUM constant exists."""
+        from praisonai.cli.branding import LOGO_MEDIUM
+        assert len(LOGO_MEDIUM) > 0
+    
+    def test_logo_small_exists(self):
+        """Test LOGO_SMALL constant exists and has branding."""
+        from praisonai.cli.branding import LOGO_SMALL
+        assert len(LOGO_SMALL) > 0
+        assert "Praison AI" in LOGO_SMALL
+    
+    def test_logo_minimal_exists(self):
+        """Test LOGO_MINIMAL constant exists and has branding."""
+        from praisonai.cli.branding import LOGO_MINIMAL
+        assert LOGO_MINIMAL == "Praison AI"
+
+
+class TestGetLogo:
+    """Tests for get_logo function."""
+    
+    def test_get_logo_wide_terminal(self):
+        """Test get_logo returns LOGO_LARGE for wide terminals."""
+        from praisonai.cli.branding import get_logo, LOGO_LARGE
+        assert get_logo(100) == LOGO_LARGE
+        assert get_logo(80) == LOGO_LARGE
+        assert get_logo(75) == LOGO_LARGE
+    
+    def test_get_logo_medium_terminal(self):
+        """Test get_logo returns LOGO_MEDIUM for medium terminals."""
+        from praisonai.cli.branding import get_logo, LOGO_MEDIUM
+        assert get_logo(74) == LOGO_MEDIUM
+        assert get_logo(50) == LOGO_MEDIUM
+        assert get_logo(40) == LOGO_MEDIUM
+    
+    def test_get_logo_narrow_terminal(self):
+        """Test get_logo returns LOGO_SMALL for narrow terminals."""
+        from praisonai.cli.branding import get_logo, LOGO_SMALL
+        assert get_logo(39) == LOGO_SMALL
+        assert get_logo(30) == LOGO_SMALL
+        assert get_logo(20) == LOGO_SMALL
+
+
+class TestGetVersion:
+    """Tests for get_version function."""
+    
+    def test_get_version_returns_string(self):
+        """Test get_version returns a string."""
+        from praisonai.cli.branding import get_version
+        version = get_version()
+        assert isinstance(version, str)
+        assert len(version) > 0
+    
+    def test_get_version_format(self):
+        """Test get_version returns a valid version format."""
+        from praisonai.cli.branding import get_version
+        version = get_version()
+        # Should be in format X.Y.Z or similar
+        parts = version.split(".")
+        assert len(parts) >= 2
+
+
+class TestGetBanner:
+    """Tests for get_banner function."""
+    
+    def test_get_banner_includes_logo(self):
+        """Test get_banner includes logo."""
+        from praisonai.cli.branding import get_banner
+        banner = get_banner(100)
+        assert "██" in banner or "Praison" in banner
+    
+    def test_get_banner_includes_version(self):
+        """Test get_banner includes version when requested."""
+        from praisonai.cli.branding import get_banner, get_version
+        banner = get_banner(100, show_version=True)
+        version = get_version()
+        assert version in banner
+    
+    def test_get_banner_includes_model(self):
+        """Test get_banner includes model when provided."""
+        from praisonai.cli.branding import get_banner
+        banner = get_banner(100, model="gpt-4o")
+        assert "gpt-4o" in banner
+
+
+class TestGetWelcomeTips:
+    """Tests for get_welcome_tips function."""
+    
+    def test_get_welcome_tips_content(self):
+        """Test get_welcome_tips returns helpful tips."""
+        from praisonai.cli.branding import get_welcome_tips
+        tips = get_welcome_tips()
+        assert "Enter" in tips or "help" in tips
+        assert len(tips) > 0

src/praisonai/tests/unit/cli/test_csv_test_runner.py

@@ -0,0 +1,343 @@
+"""
+Tests for CSV Test Runner.
+
+Tests CSV parsing, test execution, assertions, and reporting.
+"""
+
+import tempfile
+from pathlib import Path
+
+
+
+class TestCSVSchema:
+    """Tests for CSV schema definition."""
+    
+    def test_schema_has_required_fields(self):
+        """Test that schema defines required fields."""
+        from praisonai.cli.features.csv_test_runner import CSV_SCHEMA
+        
+        assert "id" in CSV_SCHEMA
+        assert "name" in CSV_SCHEMA
+        assert "prompts" in CSV_SCHEMA
+        
+        assert CSV_SCHEMA["id"]["required"] is True
+        assert CSV_SCHEMA["name"]["required"] is True
+        assert CSV_SCHEMA["prompts"]["required"] is True
+    
+    def test_schema_has_optional_fields(self):
+        """Test that schema defines optional fields."""
+        from praisonai.cli.features.csv_test_runner import CSV_SCHEMA
+        
+        optional_fields = [
+            "description", "mode", "workspace_fixture",
+            "expected_tools", "forbidden_tools", "expected_files",
+            "expected_response", "judge_rubric", "judge_threshold",
+            "judge_model", "timeout", "retries", "skip_if",
+            "agents", "workflow",
+        ]
+        
+        for field in optional_fields:
+            assert field in CSV_SCHEMA
+            assert CSV_SCHEMA[field]["required"] is False
+
+
+class TestTestCase:
+    """Tests for TestCase dataclass."""
+    
+    def test_default_values(self):
+        """Test default test case values."""
+        from praisonai.cli.features.csv_test_runner import TestCase
+        
+        tc = TestCase(id="test1", name="Test 1", prompts=["Hello"])
+        
+        assert tc.mode == "headless"
+        assert tc.workspace_fixture == "empty"
+        assert tc.expected_tools == []
+        assert tc.forbidden_tools == []
+        assert tc.judge_threshold == 7.0
+        assert tc.timeout == 60
+    
+    def test_should_skip_no_condition(self):
+        """Test should_skip with no condition."""
+        from praisonai.cli.features.csv_test_runner import TestCase
+        
+        tc = TestCase(id="test1", name="Test 1", prompts=["Hello"])
+        assert tc.should_skip() is None
+    
+    def test_should_skip_no_openai_key(self):
+        """Test should_skip with no_openai_key condition."""
+        import os
+        from praisonai.cli.features.csv_test_runner import TestCase
+        
+        tc = TestCase(
+            id="test1", name="Test 1", prompts=["Hello"],
+            skip_if="no_openai_key"
+        )
+        
+        # Save and clear key
+        original = os.environ.get("OPENAI_API_KEY")
+        if "OPENAI_API_KEY" in os.environ:
+            del os.environ["OPENAI_API_KEY"]
+        
+        try:
+            result = tc.should_skip()
+            assert result is not None
+            assert "OPENAI_API_KEY" in result
+        finally:
+            if original:
+                os.environ["OPENAI_API_KEY"] = original
+
+
+class TestTestResult:
+    """Tests for TestResult dataclass."""
+    
+    def test_to_dict(self):
+        """Test result serialization."""
+        from praisonai.cli.features.csv_test_runner import TestResult
+        
+        result = TestResult(
+            test_id="test1",
+            test_name="Test 1",
+            status="passed",
+            duration=1.5,
+            tool_calls=["read_file"],
+            response="Hello",
+        )
+        
+        d = result.to_dict()
+        assert d["test_id"] == "test1"
+        assert d["status"] == "passed"
+        assert d["duration"] == 1.5
+        assert d["tool_calls"] == ["read_file"]
+
+
+class TestTestSummary:
+    """Tests for TestSummary dataclass."""
+    
+    def test_to_dict(self):
+        """Test summary serialization."""
+        from praisonai.cli.features.csv_test_runner import TestSummary
+        
+        summary = TestSummary(
+            total=10,
+            passed=8,
+            failed=1,
+            skipped=1,
+            errors=0,
+            duration=30.0,
+        )
+        
+        d = summary.to_dict()
+        assert d["total"] == 10
+        assert d["passed"] == 8
+        assert d["pass_rate"] == 0.8
+    
+    def test_pass_rate_zero_total(self):
+        """Test pass rate with zero total."""
+        from praisonai.cli.features.csv_test_runner import TestSummary
+        
+        summary = TestSummary(total=0)
+        d = summary.to_dict()
+        assert d["pass_rate"] == 0
+
+
+class TestParseCSV:
+    """Tests for parse_csv function."""
+    
+    def test_parse_simple_csv(self):
+        """Test parsing simple CSV."""
+        from praisonai.cli.features.csv_test_runner import parse_csv
+        
+        csv_content = """id,name,prompts
+test1,Test One,Hello world
+test2,Test Two,Goodbye world"""
+        
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".csv", delete=False) as f:
+            f.write(csv_content)
+            f.flush()
+            
+            test_cases = parse_csv(Path(f.name))
+        
+        assert len(test_cases) == 2
+        assert test_cases[0].id == "test1"
+        assert test_cases[0].name == "Test One"
+        assert test_cases[0].prompts == ["Hello world"]
+    
+    def test_parse_json_prompts(self):
+        """Test parsing CSV with JSON prompts array."""
+        from praisonai.cli.features.csv_test_runner import parse_csv
+        
+        csv_content = '''id,name,prompts
+test1,Multi-step,"[""Step 1"", ""Step 2""]"'''
+        
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".csv", delete=False) as f:
+            f.write(csv_content)
+            f.flush()
+            
+            test_cases = parse_csv(Path(f.name))
+        
+        assert len(test_cases) == 1
+        assert test_cases[0].prompts == ["Step 1", "Step 2"]
+    
+    def test_parse_expected_tools(self):
+        """Test parsing expected_tools column."""
+        from praisonai.cli.features.csv_test_runner import parse_csv
+        
+        csv_content = """id,name,prompts,expected_tools
+test1,Tool Test,Hello,"read_file,write_file"
+"""
+        
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".csv", delete=False) as f:
+            f.write(csv_content)
+            f.flush()
+            
+            test_cases = parse_csv(Path(f.name))
+        
+        assert test_cases[0].expected_tools == ["read_file", "write_file"]
+    
+    def test_parse_expected_files_json(self):
+        """Test parsing expected_files JSON column."""
+        from praisonai.cli.features.csv_test_runner import parse_csv
+        
+        csv_content = '''id,name,prompts,expected_files
+test1,File Test,Hello,"{""test.py"": ""print""}"
+'''
+        
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".csv", delete=False) as f:
+            f.write(csv_content)
+            f.flush()
+            
+            test_cases = parse_csv(Path(f.name))
+        
+        assert test_cases[0].expected_files == {"test.py": "print"}
+    
+    def test_parse_agents_json(self):
+        """Test parsing agents JSON column."""
+        from praisonai.cli.features.csv_test_runner import parse_csv
+        
+        csv_content = '''id,name,prompts,agents
+test1,Multi-agent,Hello,"[{""name"": ""Agent1""}]"
+'''
+        
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".csv", delete=False) as f:
+            f.write(csv_content)
+            f.flush()
+            
+            test_cases = parse_csv(Path(f.name))
+        
+        assert test_cases[0].agents == [{"name": "Agent1"}]
+    
+    def test_parse_numeric_fields(self):
+        """Test parsing numeric fields."""
+        from praisonai.cli.features.csv_test_runner import parse_csv
+        
+        csv_content = """id,name,prompts,judge_threshold,timeout,retries
+test1,Numeric,Hello,8.5,120,2
+"""
+        
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".csv", delete=False) as f:
+            f.write(csv_content)
+            f.flush()
+            
+            test_cases = parse_csv(Path(f.name))
+        
+        assert test_cases[0].judge_threshold == 8.5
+        assert test_cases[0].timeout == 120
+        assert test_cases[0].retries == 2
+
+
+class TestGenerateCSVTemplate:
+    """Tests for generate_csv_template function."""
+    
+    def test_generates_valid_csv(self):
+        """Test that template generates valid CSV."""
+        from praisonai.cli.features.csv_test_runner import generate_csv_template
+        
+        csv_content = generate_csv_template()
+        
+        assert "id,name,prompts" in csv_content
+        assert "smoke_01" in csv_content
+        assert "tools_01" in csv_content
+    
+    def test_writes_to_file(self):
+        """Test that template can be written to file."""
+        from praisonai.cli.features.csv_test_runner import generate_csv_template
+        
+        with tempfile.NamedTemporaryFile(mode="w", suffix=".csv", delete=False) as f:
+            output_path = Path(f.name)
+        
+        generate_csv_template(output_path)
+        
+        assert output_path.exists()
+        content = output_path.read_text()
+        assert "id,name,prompts" in content
+
+
+class TestCSVTestRunner:
+    """Tests for CSVTestRunner class."""
+    
+    def test_initialization(self):
+        """Test runner initialization."""
+        from praisonai.cli.features.csv_test_runner import CSVTestRunner
+        
+        with tempfile.NamedTemporaryFile(suffix=".csv", delete=False) as f:
+            f.write(b"id,name,prompts\n")
+            csv_path = Path(f.name)
+        
+        runner = CSVTestRunner(
+            csv_path=csv_path,
+            model="gpt-4o",
+            judge_model="gpt-4o-mini",
+            fail_fast=True,
+            no_judge=True,
+        )
+        
+        assert runner.model == "gpt-4o"
+        assert runner.judge_model == "gpt-4o-mini"
+        assert runner.fail_fast is True
+        assert runner.no_judge is True
+    
+    def test_parse_judge_score_standard(self):
+        """Test parsing standard judge score format."""
+        from praisonai.cli.features.csv_test_runner import CSVTestRunner
+        
+        with tempfile.NamedTemporaryFile(suffix=".csv", delete=False) as f:
+            f.write(b"id,name,prompts\n")
+            csv_path = Path(f.name)
+        
+        runner = CSVTestRunner(csv_path=csv_path)
+        
+        response = "SCORE: 8\nREASONING: Good response"
+        score = runner._parse_judge_score(response)
+        
+        assert score == 8.0
+    
+    def test_parse_judge_score_decimal(self):
+        """Test parsing decimal judge score."""
+        from praisonai.cli.features.csv_test_runner import CSVTestRunner
+        
+        with tempfile.NamedTemporaryFile(suffix=".csv", delete=False) as f:
+            f.write(b"id,name,prompts\n")
+            csv_path = Path(f.name)
+        
+        runner = CSVTestRunner(csv_path=csv_path)
+        
+        response = "SCORE: 7.5\nREASONING: Decent response"
+        score = runner._parse_judge_score(response)
+        
+        assert score == 7.5
+    
+    def test_parse_judge_score_no_match(self):
+        """Test parsing when no score found."""
+        from praisonai.cli.features.csv_test_runner import CSVTestRunner
+        
+        with tempfile.NamedTemporaryFile(suffix=".csv", delete=False) as f:
+            f.write(b"id,name,prompts\n")
+            csv_path = Path(f.name)
+        
+        runner = CSVTestRunner(csv_path=csv_path)
+        
+        response = "This response has no score"
+        score = runner._parse_judge_score(response)
+        
+        assert score is None

src/praisonai/tests/unit/cli/test_headless_interactive_core.py

@@ -0,0 +1,309 @@
+"""
+Tests for Headless Interactive Core Executor.
+
+Tests that the headless executor uses the same interactive core pipeline
+as the TUI (InteractiveRuntime, get_interactive_tools, etc.).
+"""
+
+from unittest.mock import MagicMock
+
+import pytest
+
+
+class TestHeadlessConfig:
+    """Tests for HeadlessConfig dataclass."""
+    
+    def test_default_config(self):
+        """Test default configuration values."""
+        from praisonai.cli.features.interactive_core_headless import HeadlessConfig
+        
+        config = HeadlessConfig()
+        assert config.model == "gpt-4o-mini"
+        assert config.approval_mode == "auto"
+        assert config.enable_acp is True
+        assert config.enable_lsp is True
+        assert config.enable_basic is True
+        assert config.timeout == 60
+    
+    def test_custom_config(self):
+        """Test custom configuration values."""
+        from praisonai.cli.features.interactive_core_headless import HeadlessConfig
+        
+        config = HeadlessConfig(
+            workspace="/tmp/test",
+            model="gpt-4o",
+            approval_mode="manual",
+            timeout=120,
+        )
+        assert config.workspace == "/tmp/test"
+        assert config.model == "gpt-4o"
+        assert config.approval_mode == "manual"
+        assert config.timeout == 120
+    
+    def test_default_agent_created(self):
+        """Test that default agent is created if none provided."""
+        from praisonai.cli.features.interactive_core_headless import HeadlessConfig
+        
+        config = HeadlessConfig()
+        assert len(config.agents) == 1
+        assert config.agents[0].name == "HeadlessAgent"
+
+
+class TestAgentConfig:
+    """Tests for AgentConfig dataclass."""
+    
+    def test_default_values(self):
+        """Test default agent config values."""
+        from praisonai.cli.features.interactive_core_headless import AgentConfig
+        
+        config = AgentConfig()
+        assert config.name == "HeadlessAgent"
+        assert config.role == "assistant"
+        assert config.llm == "gpt-4o-mini"
+    
+    def test_custom_values(self):
+        """Test custom agent config values."""
+        from praisonai.cli.features.interactive_core_headless import AgentConfig
+        
+        config = AgentConfig(
+            name="TestAgent",
+            instructions="Test instructions",
+            role="tester",
+            llm="gpt-4o",
+        )
+        assert config.name == "TestAgent"
+        assert config.instructions == "Test instructions"
+        assert config.role == "tester"
+        assert config.llm == "gpt-4o"
+
+
+class TestToolCallTrace:
+    """Tests for ToolCallTrace dataclass."""
+    
+    def test_to_dict(self):
+        """Test trace serialization."""
+        from praisonai.cli.features.interactive_core_headless import ToolCallTrace
+        
+        trace = ToolCallTrace(
+            tool_name="test_tool",
+            args=("arg1",),
+            kwargs={"key": "value"},
+            result="result",
+            success=True,
+            error=None,
+            duration=0.5,
+            timestamp="2024-01-01T00:00:00",
+        )
+        
+        d = trace.to_dict()
+        assert d["tool_name"] == "test_tool"
+        assert d["success"] is True
+        assert d["duration"] == 0.5
+
+
+class TestHeadlessExecutionResult:
+    """Tests for HeadlessExecutionResult dataclass."""
+    
+    def test_to_dict(self):
+        """Test result serialization."""
+        from praisonai.cli.features.interactive_core_headless import (
+            HeadlessExecutionResult,
+        )
+        
+        result = HeadlessExecutionResult(
+            success=True,
+            responses=["Hello"],
+            tool_trace=[],
+            duration=1.0,
+            transcript=[{"role": "user", "content": "Hi"}],
+        )
+        
+        d = result.to_dict()
+        assert d["success"] is True
+        assert d["responses"] == ["Hello"]
+        assert d["duration"] == 1.0
+
+
+class TestHeadlessInteractiveCore:
+    """Tests for HeadlessInteractiveCore class."""
+    
+    def test_initialization(self):
+        """Test executor initialization."""
+        from praisonai.cli.features.interactive_core_headless import (
+            HeadlessConfig,
+            HeadlessInteractiveCore,
+        )
+        
+        config = HeadlessConfig(workspace="/tmp/test")
+        executor = HeadlessInteractiveCore(config)
+        
+        assert executor.config.workspace == "/tmp/test"
+        assert executor._initialized is False
+        assert executor._tools == []
+    
+    def test_wrap_tool_for_trace(self):
+        """Test that tool wrapping captures traces."""
+        from praisonai.cli.features.interactive_core_headless import (
+            HeadlessConfig,
+            HeadlessInteractiveCore,
+        )
+        
+        config = HeadlessConfig()
+        executor = HeadlessInteractiveCore(config)
+        
+        def sample_tool(x):
+            return x * 2
+        
+        wrapped = executor._wrap_tool_for_trace(sample_tool)
+        result = wrapped(5)
+        
+        assert result == 10
+        assert len(executor._tool_trace) == 1
+        assert executor._tool_trace[0].tool_name == "sample_tool"
+        assert executor._tool_trace[0].success is True
+    
+    def test_wrap_tool_captures_errors(self):
+        """Test that tool wrapping captures errors."""
+        from praisonai.cli.features.interactive_core_headless import (
+            HeadlessConfig,
+            HeadlessInteractiveCore,
+        )
+        
+        config = HeadlessConfig()
+        executor = HeadlessInteractiveCore(config)
+        
+        def failing_tool():
+            raise ValueError("Test error")
+        
+        wrapped = executor._wrap_tool_for_trace(failing_tool)
+        
+        with pytest.raises(ValueError):
+            wrapped()
+        
+        assert len(executor._tool_trace) == 1
+        assert executor._tool_trace[0].success is False
+        assert "Test error" in executor._tool_trace[0].error
+    
+    def test_get_agent_for_prompt_single_agent(self):
+        """Test agent selection with single agent."""
+        from praisonai.cli.features.interactive_core_headless import (
+            HeadlessConfig,
+            HeadlessInteractiveCore,
+        )
+        
+        config = HeadlessConfig()
+        executor = HeadlessInteractiveCore(config)
+        executor._agents = [MagicMock()]  # Single agent
+        
+        assert executor._get_agent_for_prompt(0, "test") == 0
+        assert executor._get_agent_for_prompt(1, "test") == 0
+    
+    def test_get_agent_for_prompt_round_robin(self):
+        """Test round-robin agent selection."""
+        from praisonai.cli.features.interactive_core_headless import (
+            HeadlessConfig,
+            HeadlessInteractiveCore,
+            AgentConfig,
+        )
+        
+        config = HeadlessConfig(
+            agents=[AgentConfig(name="A"), AgentConfig(name="B")],
+            workflow={"routing": "round_robin"},
+        )
+        executor = HeadlessInteractiveCore(config)
+        executor._agents = [MagicMock(), MagicMock()]
+        
+        assert executor._get_agent_for_prompt(0, "test") == 0
+        assert executor._get_agent_for_prompt(1, "test") == 1
+        assert executor._get_agent_for_prompt(2, "test") == 0
+    
+    def test_clear_trace(self):
+        """Test clearing tool trace."""
+        from praisonai.cli.features.interactive_core_headless import (
+            HeadlessConfig,
+            HeadlessInteractiveCore,
+            ToolCallTrace,
+        )
+        
+        config = HeadlessConfig()
+        executor = HeadlessInteractiveCore(config)
+        executor._tool_trace = [
+            ToolCallTrace("t", (), {}, None, True, None, 0, "")
+        ]
+        
+        executor.clear_trace()
+        assert executor._tool_trace == []
+
+
+class TestRunHeadlessFunction:
+    """Tests for run_headless convenience function."""
+    
+    def test_function_creates_executor(self):
+        """Test that run_headless creates executor with correct config."""
+        # We can't easily test the full run without mocking Agent
+        # Just verify the function exists and has correct signature
+        from praisonai.cli.features.interactive_core_headless import run_headless
+        
+        import inspect
+        sig = inspect.signature(run_headless)
+        params = list(sig.parameters.keys())
+        
+        assert "prompts" in params
+        assert "workspace" in params
+        assert "model" in params
+        assert "approval_mode" in params
+        assert "agents" in params
+        assert "workflow" in params
+
+
+class TestRedactSensitive:
+    """Tests for _redact_sensitive function."""
+    
+    def test_redacts_api_key(self):
+        """Test that API keys are redacted."""
+        from praisonai.cli.features.interactive_core_headless import _redact_sensitive
+        
+        data = {"api_key": "secret123", "name": "test"}
+        result = _redact_sensitive(data)
+        
+        assert result["api_key"] == "[REDACTED]"
+        assert result["name"] == "test"
+    
+    def test_redacts_nested(self):
+        """Test that nested sensitive data is redacted."""
+        from praisonai.cli.features.interactive_core_headless import _redact_sensitive
+        
+        data = {"config": {"password": "secret", "user": "admin"}}
+        result = _redact_sensitive(data)
+        
+        assert result["config"]["password"] == "[REDACTED]"
+        assert result["config"]["user"] == "admin"
+    
+    def test_truncates_long_strings(self):
+        """Test that long strings are truncated."""
+        from praisonai.cli.features.interactive_core_headless import _redact_sensitive
+        
+        long_string = "x" * 200
+        result = _redact_sensitive(long_string)
+        
+        assert len(result) < 200
+        assert "[truncated]" in result
+
+
+class TestTruncate:
+    """Tests for _truncate function."""
+    
+    def test_short_string_unchanged(self):
+        """Test that short strings are unchanged."""
+        from praisonai.cli.features.interactive_core_headless import _truncate
+        
+        result = _truncate("hello", 100)
+        assert result == "hello"
+    
+    def test_long_string_truncated(self):
+        """Test that long strings are truncated."""
+        from praisonai.cli.features.interactive_core_headless import _truncate
+        
+        result = _truncate("x" * 100, 50)
+        assert len(result) < 100
+        assert "[truncated]" in result

src/praisonai/tests/unit/cli/test_interactive_test_harness.py

@@ -0,0 +1,344 @@
+"""
+Tests for Interactive Test Harness.
+
+Tests workspace isolation, fixtures, tool trace, and artifact generation.
+"""
+
+import tempfile
+from pathlib import Path
+
+
+class TestWorkspaceFixture:
+    """Tests for WorkspaceFixture dataclass."""
+    
+    def test_empty_fixture(self):
+        """Test empty workspace fixture."""
+        from praisonai.cli.features.interactive_test_harness import WorkspaceFixture
+        
+        fixture = WorkspaceFixture.empty()
+        assert fixture.name == "empty"
+        assert fixture.files == {}
+        assert fixture.directories == []
+    
+    def test_seeded_fixture(self):
+        """Test seeded workspace fixture."""
+        from praisonai.cli.features.interactive_test_harness import WorkspaceFixture
+        
+        fixture = WorkspaceFixture.seeded()
+        assert fixture.name == "seeded"
+        assert "README.md" in fixture.files
+        assert "main.py" in fixture.files
+        assert "src" in fixture.directories
+    
+    def test_git_fixture(self):
+        """Test git workspace fixture."""
+        from praisonai.cli.features.interactive_test_harness import WorkspaceFixture
+        
+        fixture = WorkspaceFixture.git()
+        assert fixture.name == "git"
+        assert ".gitignore" in fixture.files
+    
+    def test_python_project_fixture(self):
+        """Test Python project workspace fixture."""
+        from praisonai.cli.features.interactive_test_harness import WorkspaceFixture
+        
+        fixture = WorkspaceFixture.python_project()
+        assert fixture.name == "python_project"
+        assert "main.py" in fixture.files
+        assert "utils.py" in fixture.files
+        assert "def main" in fixture.files["main.py"]
+
+
+class TestBuiltinFixtures:
+    """Tests for BUILTIN_FIXTURES registry."""
+    
+    def test_all_fixtures_registered(self):
+        """Test that all fixtures are registered."""
+        from praisonai.cli.features.interactive_test_harness import BUILTIN_FIXTURES
+        
+        assert "empty" in BUILTIN_FIXTURES
+        assert "seeded" in BUILTIN_FIXTURES
+        assert "git" in BUILTIN_FIXTURES
+        assert "python_project" in BUILTIN_FIXTURES
+    
+    def test_fixtures_are_callable(self):
+        """Test that fixtures are callable factories."""
+        from praisonai.cli.features.interactive_test_harness import BUILTIN_FIXTURES
+        
+        for name, factory in BUILTIN_FIXTURES.items():
+            fixture = factory()
+            assert fixture.name == name
+
+
+class TestTestArtifacts:
+    """Tests for TestArtifacts dataclass."""
+    
+    def test_save_creates_files(self):
+        """Test that save creates artifact files."""
+        from praisonai.cli.features.interactive_test_harness import TestArtifacts
+        
+        with tempfile.TemporaryDirectory() as tmpdir:
+            artifacts_dir = Path(tmpdir) / "artifacts"
+            
+            artifacts = TestArtifacts(
+                transcript=[{"role": "user", "content": "Hello"}],
+                tool_trace=[{"tool": "test", "success": True}],
+                result={"status": "passed"},
+            )
+            
+            artifacts.save(artifacts_dir)
+            
+            assert (artifacts_dir / "transcript.txt").exists()
+            assert (artifacts_dir / "tool_trace.jsonl").exists()
+            assert (artifacts_dir / "result.json").exists()
+    
+    def test_save_with_workspace_snapshot(self):
+        """Test saving with workspace snapshot."""
+        from praisonai.cli.features.interactive_test_harness import TestArtifacts
+        
+        with tempfile.TemporaryDirectory() as tmpdir:
+            artifacts_dir = Path(tmpdir) / "artifacts"
+            
+            artifacts = TestArtifacts(
+                transcript=[],
+                tool_trace=[],
+                result={},
+                workspace_snapshot={"test.py": "print('hello')"},
+            )
+            
+            artifacts.save(artifacts_dir)
+            
+            snapshot_file = artifacts_dir / "workspace" / "test.py"
+            assert snapshot_file.exists()
+            assert snapshot_file.read_text() == "print('hello')"
+    
+    def test_save_with_judge_result(self):
+        """Test saving with judge result."""
+        from praisonai.cli.features.interactive_test_harness import TestArtifacts
+        
+        with tempfile.TemporaryDirectory() as tmpdir:
+            artifacts_dir = Path(tmpdir) / "artifacts"
+            
+            artifacts = TestArtifacts(
+                transcript=[],
+                tool_trace=[],
+                result={},
+                judge_result={"score": 8, "passed": True},
+            )
+            
+            artifacts.save(artifacts_dir)
+            
+            assert (artifacts_dir / "judge_result.json").exists()
+
+
+class TestInteractiveTestHarness:
+    """Tests for InteractiveTestHarness class."""
+    
+    def test_initialization_creates_workspace(self):
+        """Test that initialization creates workspace."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        harness = InteractiveTestHarness()
+        
+        assert harness.workspace.exists()
+        assert harness._workspace_created is True
+        
+        harness.cleanup()
+    
+    def test_initialization_with_custom_workspace(self):
+        """Test initialization with custom workspace."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        with tempfile.TemporaryDirectory() as tmpdir:
+            workspace = Path(tmpdir) / "custom"
+            workspace.mkdir()
+            
+            harness = InteractiveTestHarness(workspace=workspace)
+            
+            assert harness.workspace == workspace
+            assert harness._workspace_created is False
+    
+    def test_setup_workspace_empty(self):
+        """Test setting up empty workspace."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        harness = InteractiveTestHarness()
+        harness.setup_workspace("empty")
+        
+        assert harness.workspace.exists()
+        # Empty workspace should have no files
+        files = list(harness.workspace.iterdir())
+        assert len(files) == 0
+        
+        harness.cleanup()
+    
+    def test_setup_workspace_seeded(self):
+        """Test setting up seeded workspace."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        harness = InteractiveTestHarness()
+        harness.setup_workspace("seeded")
+        
+        assert (harness.workspace / "README.md").exists()
+        assert (harness.workspace / "main.py").exists()
+        assert (harness.workspace / "src").is_dir()
+        
+        harness.cleanup()
+    
+    def test_verify_files_exists(self):
+        """Test verifying file existence."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        harness = InteractiveTestHarness()
+        harness.setup_workspace("empty")
+        
+        # Create a test file
+        test_file = harness.workspace / "test.py"
+        test_file.write_text("print('hello')")
+        
+        results = harness.verify_files({"test.py": ""})
+        assert results["test.py"] is True
+        
+        results = harness.verify_files({"nonexistent.py": ""})
+        assert results["nonexistent.py"] is False
+        
+        harness.cleanup()
+    
+    def test_verify_files_content_pattern(self):
+        """Test verifying file content with pattern."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        harness = InteractiveTestHarness()
+        harness.setup_workspace("empty")
+        
+        test_file = harness.workspace / "test.py"
+        test_file.write_text("x = 42\nprint(x)")
+        
+        # Pattern matches
+        results = harness.verify_files({"test.py": "x.*=.*42"})
+        assert results["test.py"] is True
+        
+        # Pattern doesn't match
+        results = harness.verify_files({"test.py": "y.*=.*99"})
+        assert results["test.py"] is False
+        
+        harness.cleanup()
+    
+    def test_verify_tool_calls_expected(self):
+        """Test verifying expected tool calls."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        from unittest.mock import MagicMock
+        
+        harness = InteractiveTestHarness()
+        harness._executor = MagicMock()
+        harness._executor.get_tools_called.return_value = ["read_file", "write_file"]
+        
+        results = harness.verify_tool_calls(expected_tools=["read_file"])
+        assert results["passed"] is True
+        assert results["expected_passed"] is True
+        
+        results = harness.verify_tool_calls(expected_tools=["delete_file"])
+        assert results["passed"] is False
+        assert "delete_file" in results["missing_tools"]
+        
+        harness.cleanup()
+    
+    def test_verify_tool_calls_forbidden(self):
+        """Test verifying forbidden tool calls."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        from unittest.mock import MagicMock
+        
+        harness = InteractiveTestHarness()
+        harness._executor = MagicMock()
+        harness._executor.get_tools_called.return_value = ["read_file", "delete_file"]
+        
+        results = harness.verify_tool_calls(forbidden_tools=["delete_file"])
+        assert results["passed"] is False
+        assert results["forbidden_passed"] is False
+        assert "delete_file" in results["forbidden_called"]
+        
+        harness.cleanup()
+    
+    def test_verify_response_pattern(self):
+        """Test verifying response with pattern."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        harness = InteractiveTestHarness()
+        
+        assert harness.verify_response("The answer is 42", "42") is True
+        assert harness.verify_response("The answer is 42", "99") is False
+        assert harness.verify_response("Hello World", "hello") is True  # Case insensitive
+        
+        harness.cleanup()
+    
+    def test_snapshot_workspace(self):
+        """Test creating workspace snapshot."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        harness = InteractiveTestHarness()
+        harness.setup_workspace("seeded")
+        
+        snapshot = harness.snapshot_workspace()
+        
+        assert "README.md" in snapshot
+        assert "main.py" in snapshot
+        assert "Test Project" in snapshot["README.md"]
+        
+        harness.cleanup()
+    
+    def test_context_manager(self):
+        """Test using harness as context manager."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        with InteractiveTestHarness() as harness:
+            harness.setup_workspace("empty")
+            workspace_path = harness.workspace
+            assert workspace_path.exists()
+        
+        # Workspace should be cleaned up
+        assert not workspace_path.exists()
+    
+    def test_keep_workspace(self):
+        """Test keeping workspace after cleanup."""
+        from praisonai.cli.features.interactive_test_harness import InteractiveTestHarness
+        
+        harness = InteractiveTestHarness(keep_workspace=True)
+        harness.setup_workspace("empty")
+        workspace_path = harness.workspace
+        
+        harness.cleanup()
+        
+        # Workspace should still exist
+        assert workspace_path.exists()
+        
+        # Manual cleanup
+        import shutil
+        shutil.rmtree(workspace_path)
+
+
+class TestCreateTestHarness:
+    """Tests for create_test_harness function."""
+    
+    def test_creates_harness_with_fixture(self):
+        """Test creating harness with fixture."""
+        from praisonai.cli.features.interactive_test_harness import create_test_harness
+        
+        harness = create_test_harness(fixture="seeded")
+        
+        assert (harness.workspace / "README.md").exists()
+        
+        harness.cleanup()
+    
+    def test_creates_harness_with_custom_workspace(self):
+        """Test creating harness with custom workspace."""
+        from praisonai.cli.features.interactive_test_harness import create_test_harness
+        
+        with tempfile.TemporaryDirectory() as tmpdir:
+            workspace = Path(tmpdir) / "custom"
+            
+            harness = create_test_harness(
+                fixture="empty",
+                workspace=workspace,
+            )
+            
+            assert harness.workspace == workspace

src/praisonai/tests/unit/cli/test_praison_io.py

@@ -0,0 +1,406 @@
+"""
+Tests for PraisonIO - Interactive CLI I/O handler.
+"""
+
+from unittest.mock import patch
+
+
+class TestIOConfig:
+    """Tests for IOConfig dataclass."""
+    
+    def test_default_values(self):
+        """Test default configuration values."""
+        from praisonai.cli.interactive.praison_io import IOConfig
+        
+        config = IOConfig()
+        assert config.pretty is True
+        assert config.user_input_color == "#00cc00"
+        assert config.multiline_mode is False
+        assert config.vi_mode is False
+        assert config.enable_completions is True
+    
+    def test_custom_values(self):
+        """Test custom configuration values."""
+        from praisonai.cli.interactive.praison_io import IOConfig
+        
+        config = IOConfig(
+            pretty=False,
+            user_input_color="#ff0000",
+            multiline_mode=True,
+            vi_mode=True,
+        )
+        assert config.pretty is False
+        assert config.user_input_color == "#ff0000"
+        assert config.multiline_mode is True
+        assert config.vi_mode is True
+
+
+class TestPraisonCompleter:
+    """Tests for PraisonCompleter."""
+    
+    def test_init_empty(self):
+        """Test initialization with no commands."""
+        from praisonai.cli.interactive.praison_io import PraisonCompleter
+        
+        completer = PraisonCompleter()
+        assert completer.commands == {}
+        assert len(completer.files) == 0
+        assert len(completer.symbols) == 0
+    
+    def test_init_with_commands(self):
+        """Test initialization with commands."""
+        from praisonai.cli.interactive.praison_io import PraisonCompleter
+        
+        commands = {"help": "Show help", "exit": "Exit"}
+        completer = PraisonCompleter(commands=commands)
+        assert completer.commands == commands
+    
+    def test_add_files(self):
+        """Test adding files."""
+        from praisonai.cli.interactive.praison_io import PraisonCompleter
+        
+        completer = PraisonCompleter()
+        completer.add_files(["file1.py", "file2.py"])
+        assert "file1.py" in completer.files
+        assert "file2.py" in completer.files
+    
+    def test_add_symbols(self):
+        """Test adding symbols."""
+        from praisonai.cli.interactive.praison_io import PraisonCompleter
+        
+        completer = PraisonCompleter()
+        completer.add_symbols(["MyClass", "my_function"])
+        assert "MyClass" in completer.symbols
+        assert "my_function" in completer.symbols
+    
+    def test_get_completions_slash_commands(self):
+        """Test completions for slash commands."""
+        from praisonai.cli.interactive.praison_io import PraisonCompleter
+        
+        commands = {"help": "Show help", "history": "Show history", "exit": "Exit"}
+        completer = PraisonCompleter(commands=commands)
+        
+        completions = completer.get_completions("/h", 2)
+        completion_texts = [c[0] for c in completions]
+        assert "/help" in completion_texts
+        assert "/history" in completion_texts
+        assert "/exit" not in completion_texts
+    
+    def test_get_completions_file_mentions(self):
+        """Test completions for @ file mentions."""
+        from praisonai.cli.interactive.praison_io import PraisonCompleter
+        
+        completer = PraisonCompleter()
+        completer.add_files(["main.py", "test.py", "README.md"])
+        
+        completions = completer.get_completions("@main", 5)
+        completion_texts = [c[0] for c in completions]
+        assert "@main.py" in completion_texts
+    
+    def test_get_completions_symbols(self):
+        """Test completions for symbols."""
+        from praisonai.cli.interactive.praison_io import PraisonCompleter
+        
+        completer = PraisonCompleter()
+        completer.add_symbols(["MyClass", "MyFunction", "other"])
+        
+        completions = completer.get_completions("My", 2)
+        completion_texts = [c[0] for c in completions]
+        assert "MyClass" in completion_texts
+        assert "MyFunction" in completion_texts
+        assert "other" not in completion_texts
+
+
+class TestMarkdownStream:
+    """Tests for MarkdownStream."""
+    
+    def test_init(self):
+        """Test initialization."""
+        from praisonai.cli.interactive.praison_io import MarkdownStream
+        
+        stream = MarkdownStream(live_window=8)
+        assert stream.live_window == 8
+        assert stream.printed == []
+        assert stream.live is None
+    
+    def test_render_to_lines_fallback(self):
+        """Test rendering without Rich."""
+        from praisonai.cli.interactive.praison_io import MarkdownStream
+        
+        stream = MarkdownStream()
+        # This tests the fallback path
+        text = "Hello\nWorld"
+        lines = stream._render_to_lines(text)
+        assert len(lines) >= 1
+
+
+class TestPraisonIO:
+    """Tests for PraisonIO."""
+    
+    def test_init_default(self):
+        """Test default initialization."""
+        from praisonai.cli.interactive.praison_io import PraisonIO
+        
+        io = PraisonIO()
+        assert io.config is not None
+        assert io.completer is not None
+    
+    def test_init_with_config(self):
+        """Test initialization with config."""
+        from praisonai.cli.interactive.praison_io import PraisonIO, IOConfig
+        
+        config = IOConfig(pretty=False)
+        io = PraisonIO(config=config)
+        assert io.config.pretty is False
+    
+    def test_add_commands(self):
+        """Test adding commands."""
+        from praisonai.cli.interactive.praison_io import PraisonIO
+        
+        io = PraisonIO()
+        io.add_commands({"test": "Test command"})
+        assert "test" in io.completer.commands
+    
+    def test_add_files(self):
+        """Test adding files."""
+        from praisonai.cli.interactive.praison_io import PraisonIO
+        
+        io = PraisonIO()
+        io.add_files(["file.py"])
+        assert "file.py" in io.completer.files
+    
+    def test_add_symbols(self):
+        """Test adding symbols."""
+        from praisonai.cli.interactive.praison_io import PraisonIO
+        
+        io = PraisonIO()
+        io.add_symbols(["MyClass"])
+        assert "MyClass" in io.completer.symbols
+    
+    @patch('builtins.print')
+    def test_rule_no_rich(self, mock_print):
+        """Test rule without Rich."""
+        from praisonai.cli.interactive.praison_io import PraisonIO, IOConfig
+        
+        config = IOConfig(pretty=False)
+        io = PraisonIO(config=config)
+        io.rule()
+        mock_print.assert_called()
+    
+    @patch('builtins.print')
+    def test_tool_output_no_rich(self, mock_print):
+        """Test tool output without Rich."""
+        from praisonai.cli.interactive.praison_io import PraisonIO, IOConfig
+        
+        config = IOConfig(pretty=False)
+        io = PraisonIO(config=config)
+        io.tool_output("Test message")
+        mock_print.assert_called_with("⚙ Test message")
+    
+    @patch('builtins.print')
+    def test_info_no_rich(self, mock_print):
+        """Test info without Rich."""
+        from praisonai.cli.interactive.praison_io import PraisonIO, IOConfig
+        
+        config = IOConfig(pretty=False)
+        io = PraisonIO(config=config)
+        io.info("Test info")
+        mock_print.assert_called_with("ℹ Test info")
+    
+    @patch('builtins.print')
+    def test_success_no_rich(self, mock_print):
+        """Test success without Rich."""
+        from praisonai.cli.interactive.praison_io import PraisonIO, IOConfig
+        
+        config = IOConfig(pretty=False)
+        io = PraisonIO(config=config)
+        io.success("Test success")
+        mock_print.assert_called_with("✓ Test success")
+
+
+class TestREPLConfig:
+    """Tests for REPLConfig."""
+    
+    def test_default_values(self):
+        """Test default configuration values."""
+        from praisonai.cli.interactive.repl import REPLConfig
+        
+        config = REPLConfig()
+        assert config.model == "gpt-4o-mini"
+        assert config.verbose is False
+        assert config.memory is False
+        assert config.tools is None
+        assert config.autonomy is True
+    
+    def test_custom_values(self):
+        """Test custom configuration values."""
+        from praisonai.cli.interactive.repl import REPLConfig
+        
+        config = REPLConfig(
+            model="gpt-4o",
+            verbose=True,
+            memory=True,
+        )
+        assert config.model == "gpt-4o"
+        assert config.verbose is True
+        assert config.memory is True
+
+
+class TestInteractiveREPL:
+    """Tests for InteractiveREPL."""
+    
+    def test_init_default(self):
+        """Test default initialization."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        assert repl.config is not None
+        assert repl.io is not None
+        assert repl._running is False
+    
+    def test_init_with_config(self):
+        """Test initialization with config."""
+        from praisonai.cli.interactive.repl import InteractiveREPL, REPLConfig
+        
+        config = REPLConfig(model="gpt-4o")
+        repl = InteractiveREPL(config=config)
+        assert repl.config.model == "gpt-4o"
+    
+    def test_handle_command_exit(self):
+        """Test handling exit command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        repl._running = True
+        
+        result = repl._handle_command("/exit")
+        assert result is True
+        assert repl._running is False
+    
+    def test_handle_command_quit(self):
+        """Test handling quit command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        repl._running = True
+        
+        result = repl._handle_command("/quit")
+        assert result is True
+        assert repl._running is False
+    
+    def test_handle_command_help(self):
+        """Test handling help command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        result = repl._handle_command("/help")
+        assert result is True
+    
+    def test_handle_command_clear(self):
+        """Test handling clear command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        repl._conversation_history = [{"role": "user", "content": "test"}]
+        
+        result = repl._handle_command("/clear")
+        assert result is True
+        assert len(repl._conversation_history) == 0
+    
+    def test_handle_command_model_show(self):
+        """Test handling model command (show)."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        result = repl._handle_command("/model")
+        assert result is True
+    
+    def test_handle_command_model_change(self):
+        """Test handling model command (change)."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        result = repl._handle_command("/model gpt-4o")
+        assert result is True
+        assert repl.config.model == "gpt-4o"
+    
+    def test_handle_command_unknown(self):
+        """Test handling unknown command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        result = repl._handle_command("/unknown")
+        assert result is True  # Still handled (shows warning)
+    
+    def test_handle_command_session(self):
+        """Test handling session command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        repl._session_id = "test-session-123"
+        
+        result = repl._handle_command("/session")
+        assert result is True
+    
+    def test_handle_command_cost(self):
+        """Test handling cost command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        repl._total_tokens = 100
+        repl._total_cost = 0.01
+        
+        result = repl._handle_command("/cost")
+        assert result is True
+    
+    def test_handle_command_history_empty(self):
+        """Test handling history command with empty history."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        result = repl._handle_command("/history")
+        assert result is True
+    
+    def test_handle_command_history_with_items(self):
+        """Test handling history command with items."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        repl._conversation_history = [
+            {"role": "user", "content": "Hello"},
+            {"role": "assistant", "content": "Hi there!"},
+        ]
+        
+        result = repl._handle_command("/history")
+        assert result is True
+    
+    def test_handle_command_compact(self):
+        """Test handling compact command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        original = repl.io.config.pretty
+        
+        result = repl._handle_command("/compact")
+        assert result is True
+        assert repl.io.config.pretty != original
+    
+    def test_handle_command_multiline(self):
+        """Test handling multiline command."""
+        from praisonai.cli.interactive.repl import InteractiveREPL
+        
+        repl = InteractiveREPL()
+        original = repl.io.config.multiline_mode
+        
+        result = repl._handle_command("/multiline")
+        assert result is True
+        assert repl.io.config.multiline_mode != original
+
+
+class TestStartInteractive:
+    """Tests for start_interactive function."""
+    
+    def test_function_exists(self):
+        """Test that start_interactive function exists."""
+        from praisonai.cli.interactive.repl import start_interactive
+        assert callable(start_interactive)

src/praisonai/tests/unit/cli/test_split_tui.py

@@ -0,0 +1,282 @@
+"""
+Tests for PraisonAI Split-Pane TUI Application.
+"""
+
+
+class TestSplitTUIConfig:
+    """Tests for SplitTUIConfig dataclass."""
+    
+    def test_default_values(self):
+        """Test default configuration values."""
+        from praisonai.cli.interactive.split_tui import SplitTUIConfig
+        
+        config = SplitTUIConfig()
+        assert config.model == "gpt-4o-mini"
+        assert config.show_logo is True
+        assert config.show_status_bar is True
+        assert config.session_id is None
+        assert config.workspace is None
+    
+    def test_custom_values(self):
+        """Test custom configuration values."""
+        from praisonai.cli.interactive.split_tui import SplitTUIConfig
+        
+        config = SplitTUIConfig(
+            model="gpt-4o",
+            show_logo=False,
+            session_id="test123",
+        )
+        assert config.model == "gpt-4o"
+        assert config.show_logo is False
+        assert config.session_id == "test123"
+
+
+class TestChatMessage:
+    """Tests for ChatMessage dataclass."""
+    
+    def test_message_creation(self):
+        """Test message creation."""
+        from praisonai.cli.interactive.split_tui import ChatMessage
+        
+        msg = ChatMessage(role="user", content="Hello")
+        assert msg.role == "user"
+        assert msg.content == "Hello"
+        assert msg.timestamp is not None
+    
+    def test_message_roles(self):
+        """Test different message roles."""
+        from praisonai.cli.interactive.split_tui import ChatMessage
+        
+        roles = ["user", "assistant", "system", "status"]
+        for role in roles:
+            msg = ChatMessage(role=role, content="test")
+            assert msg.role == role
+
+
+class TestGetLogo:
+    """Tests for get_logo function."""
+    
+    def test_logo_wide_terminal(self):
+        """Test logo for wide terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_LARGE
+        
+        logo = get_logo(100)
+        assert logo == LOGO_LARGE
+    
+    def test_logo_medium_terminal(self):
+        """Test logo for medium terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_MEDIUM
+        
+        logo = get_logo(50)
+        assert logo == LOGO_MEDIUM
+    
+    def test_logo_narrow_terminal(self):
+        """Test logo for narrow terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_SMALL
+        
+        logo = get_logo(30)
+        assert logo == LOGO_SMALL
+    
+    def test_logo_contains_praison_ai(self):
+        """Test that logo contains 'Praison AI' branding."""
+        from praisonai.cli.branding import LOGO_SMALL
+        
+        assert "Praison AI" in LOGO_SMALL
+
+
+class TestSplitTUI:
+    """Tests for SplitTUI class."""
+    
+    def test_init_default(self):
+        """Test default initialization."""
+        from praisonai.cli.interactive.split_tui import SplitTUI
+        
+        tui = SplitTUI()
+        assert tui.config is not None
+        assert tui.messages == []
+        assert tui._running is False
+        assert tui.session_id is not None
+    
+    def test_init_with_config(self):
+        """Test initialization with config."""
+        from praisonai.cli.interactive.split_tui import SplitTUI, SplitTUIConfig
+        
+        config = SplitTUIConfig(model="gpt-4o", show_logo=False)
+        tui = SplitTUI(config=config)
+        assert tui.config.model == "gpt-4o"
+        assert tui.config.show_logo is False
+    
+    def test_format_output_empty(self):
+        """Test format_output with no messages."""
+        from praisonai.cli.interactive.split_tui import SplitTUI, SplitTUIConfig
+        
+        config = SplitTUIConfig(show_logo=True)
+        tui = SplitTUI(config=config)
+        
+        output = tui._format_output()
+        # Should contain welcome message
+        assert "Type your message" in output or "gpt-4o-mini" in output
+    
+    def test_format_output_with_messages(self):
+        """Test format_output with messages."""
+        from praisonai.cli.interactive.split_tui import SplitTUI, ChatMessage
+        
+        tui = SplitTUI()
+        tui.messages = [
+            ChatMessage(role="user", content="Hello"),
+            ChatMessage(role="assistant", content="Hi there"),
+        ]
+        
+        output = tui._format_output()
+        assert "Hello" in output
+        assert "Hi there" in output
+    
+    def test_handle_command_exit(self):
+        """Test handling exit command."""
+        from praisonai.cli.interactive.split_tui import SplitTUI
+        
+        tui = SplitTUI()
+        tui._running = True
+        
+        result = tui._handle_command("/exit")
+        assert result is True
+        assert tui._running is False
+    
+    def test_handle_command_quit(self):
+        """Test handling quit command."""
+        from praisonai.cli.interactive.split_tui import SplitTUI
+        
+        tui = SplitTUI()
+        tui._running = True
+        
+        result = tui._handle_command("/quit")
+        assert result is True
+        assert tui._running is False
+    
+    def test_handle_command_help(self):
+        """Test handling help command."""
+        from praisonai.cli.interactive.split_tui import SplitTUI
+        
+        tui = SplitTUI()
+        result = tui._handle_command("/help")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert tui.messages[0].role == "system"
+        assert "/help" in tui.messages[0].content
+    
+    def test_handle_command_clear(self):
+        """Test handling clear command."""
+        from praisonai.cli.interactive.split_tui import SplitTUI, ChatMessage
+        
+        tui = SplitTUI()
+        tui.messages = [ChatMessage(role="user", content="test")]
+        
+        result = tui._handle_command("/clear")
+        assert result is True
+        assert len(tui.messages) == 0
+    
+    def test_handle_command_new(self):
+        """Test handling new command."""
+        from praisonai.cli.interactive.split_tui import SplitTUI, ChatMessage
+        
+        tui = SplitTUI()
+        old_session = tui.session_id
+        tui.messages = [ChatMessage(role="user", content="test")]
+        
+        result = tui._handle_command("/new")
+        assert result is True
+        assert len(tui.messages) == 1  # New session message
+        assert tui.session_id != old_session
+    
+    def test_handle_command_model_show(self):
+        """Test handling model command (show)."""
+        from praisonai.cli.interactive.split_tui import SplitTUI
+        
+        tui = SplitTUI()
+        result = tui._handle_command("/model")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "gpt-4o-mini" in tui.messages[0].content
+    
+    def test_handle_command_model_change(self):
+        """Test handling model command (change)."""
+        from praisonai.cli.interactive.split_tui import SplitTUI
+        
+        tui = SplitTUI()
+        result = tui._handle_command("/model gpt-4o")
+        assert result is True
+        assert tui.config.model == "gpt-4o"
+        assert tui._agent is None  # Agent should be reset
+    
+    def test_handle_command_unknown(self):
+        """Test handling unknown command."""
+        from praisonai.cli.interactive.split_tui import SplitTUI
+        
+        tui = SplitTUI()
+        result = tui._handle_command("/unknown")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "Unknown" in tui.messages[0].content
+
+
+class TestStartSplitTUI:
+    """Tests for start_split_tui function."""
+    
+    def test_function_exists(self):
+        """Test that start_split_tui function exists."""
+        from praisonai.cli.interactive.split_tui import start_split_tui
+        assert callable(start_split_tui)
+
+
+class TestASCIILogos:
+    """Tests for ASCII logo constants."""
+    
+    def test_logo_exists(self):
+        """Test LOGO constant exists and has content."""
+        from praisonai.cli.branding import LOGO_LARGE
+        assert len(LOGO_LARGE) > 0
+        assert "██" in LOGO_LARGE  # Contains block characters
+    
+    def test_logo_small_exists(self):
+        """Test LOGO_SMALL constant exists."""
+        from praisonai.cli.branding import LOGO_SMALL
+        assert len(LOGO_SMALL) > 0
+    
+    def test_logo_minimal_branding(self):
+        """Test LOGO_MINIMAL has correct branding."""
+        from praisonai.cli.branding import LOGO_MINIMAL
+        assert "Praison AI" in LOGO_MINIMAL
+
+
+class TestMessageFormatting:
+    """Tests for message formatting."""
+    
+    def test_user_message_prefix(self):
+        """Test user messages have correct prefix."""
+        from praisonai.cli.interactive.split_tui import SplitTUI, ChatMessage
+        
+        tui = SplitTUI()
+        tui.messages = [ChatMessage(role="user", content="test")]
+        
+        output = tui._format_output()
+        assert "› test" in output
+    
+    def test_assistant_message_prefix(self):
+        """Test assistant messages have correct prefix."""
+        from praisonai.cli.interactive.split_tui import SplitTUI, ChatMessage
+        
+        tui = SplitTUI()
+        tui.messages = [ChatMessage(role="assistant", content="response")]
+        
+        output = tui._format_output()
+        assert "● response" in output
+    
+    def test_status_message_prefix(self):
+        """Test status messages have correct prefix."""
+        from praisonai.cli.interactive.split_tui import SplitTUI, ChatMessage
+        
+        tui = SplitTUI()
+        tui.messages = [ChatMessage(role="status", content="thinking")]
+        
+        output = tui._format_output()
+        assert "⏳ thinking" in output

src/praisonai/tests/unit/cli/test_tui_app.py

@@ -0,0 +1,344 @@
+"""
+Tests for PraisonAI TUI Application.
+"""
+
+
+
+class TestColorScheme:
+    """Tests for ColorScheme dataclass."""
+    
+    def test_default_values(self):
+        """Test default color scheme values."""
+        from praisonai.cli.interactive.tui_app import ColorScheme
+        
+        colors = ColorScheme()
+        assert colors.primary == "#00aaff"
+        assert colors.secondary == "#888888"
+        assert colors.accent == "#00cc00"
+        assert colors.text == "#ffffff"
+        assert colors.success == "#00cc00"
+        assert colors.error == "#ff4444"
+
+
+class TestTUIConfig:
+    """Tests for TUIConfig dataclass."""
+    
+    def test_default_values(self):
+        """Test default TUI configuration values."""
+        from praisonai.cli.interactive.tui_app import TUIConfig
+        
+        config = TUIConfig()
+        assert config.model == "gpt-4o-mini"
+        assert config.show_logo is True
+        assert config.show_tips is True
+        assert config.show_status_bar is True
+        assert config.compact_mode is False
+        assert config.multiline is False
+        assert config.vi_mode is False
+    
+    def test_custom_values(self):
+        """Test custom TUI configuration values."""
+        from praisonai.cli.interactive.tui_app import TUIConfig
+        
+        config = TUIConfig(
+            model="gpt-4o",
+            show_logo=False,
+            compact_mode=True,
+        )
+        assert config.model == "gpt-4o"
+        assert config.show_logo is False
+        assert config.compact_mode is True
+
+
+class TestMessage:
+    """Tests for Message dataclass."""
+    
+    def test_message_creation(self):
+        """Test message creation."""
+        from praisonai.cli.interactive.tui_app import Message
+        
+        msg = Message(role="user", content="Hello")
+        assert msg.role == "user"
+        assert msg.content == "Hello"
+        assert msg.timestamp is not None
+        assert msg.metadata == {}
+    
+    def test_message_with_metadata(self):
+        """Test message with metadata."""
+        from praisonai.cli.interactive.tui_app import Message
+        
+        msg = Message(role="assistant", content="Hi", metadata={"tokens": 10})
+        assert msg.metadata == {"tokens": 10}
+
+
+class TestGetLogo:
+    """Tests for get_logo function."""
+    
+    def test_logo_wide_terminal(self):
+        """Test logo for wide terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_LARGE
+        
+        logo = get_logo(100)
+        assert logo == LOGO_LARGE
+    
+    def test_logo_medium_terminal(self):
+        """Test logo for medium terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_MEDIUM
+        
+        logo = get_logo(65)
+        assert logo == LOGO_MEDIUM
+    
+    def test_logo_narrow_terminal(self):
+        """Test logo for narrow terminal."""
+        from praisonai.cli.branding import get_logo, LOGO_SMALL
+        
+        logo = get_logo(30)
+        assert logo == LOGO_SMALL
+
+
+class TestPraisonTUI:
+    """Tests for PraisonTUI class."""
+    
+    def test_init_default(self):
+        """Test default initialization."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        assert tui.config is not None
+        assert tui.messages == []
+        assert tui._running is False
+        assert tui.session_id is not None
+    
+    def test_init_with_config(self):
+        """Test initialization with config."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, TUIConfig
+        
+        config = TUIConfig(model="gpt-4o", show_logo=False)
+        tui = PraisonTUI(config=config)
+        assert tui.config.model == "gpt-4o"
+        assert tui.config.show_logo is False
+    
+    def test_render_logo_disabled(self):
+        """Test render_logo when disabled."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, TUIConfig
+        
+        config = TUIConfig(show_logo=False)
+        tui = PraisonTUI(config=config)
+        
+        logo = tui._render_logo()
+        assert logo == ""
+    
+    def test_render_logo_enabled(self):
+        """Test render_logo when enabled."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, TUIConfig
+        
+        config = TUIConfig(show_logo=True, model="gpt-4o-mini")
+        tui = PraisonTUI(config=config)
+        
+        logo = tui._render_logo()
+        assert "PRAISON" in logo or "gpt-4o-mini" in logo
+    
+    def test_render_welcome(self):
+        """Test render_welcome."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, TUIConfig
+        
+        config = TUIConfig(show_logo=True, show_tips=True)
+        tui = PraisonTUI(config=config)
+        
+        welcome = tui._render_welcome()
+        assert "Type your message" in welcome or len(welcome) > 0
+    
+    def test_render_status_bar(self):
+        """Test render_status_bar."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, TUIConfig
+        
+        config = TUIConfig(show_status_bar=True, model="gpt-4o")
+        tui = PraisonTUI(config=config)
+        
+        status = tui._render_status_bar()
+        assert "gpt-4o" in status
+    
+    def test_render_status_bar_disabled(self):
+        """Test render_status_bar when disabled."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, TUIConfig
+        
+        config = TUIConfig(show_status_bar=False)
+        tui = PraisonTUI(config=config)
+        
+        status = tui._render_status_bar()
+        assert status == ""
+    
+    def test_format_message_user(self):
+        """Test format_message for user role."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, Message
+        
+        tui = PraisonTUI()
+        msg = Message(role="user", content="Hello")
+        
+        formatted = tui._format_message(msg)
+        assert "Hello" in formatted
+        assert "›" in formatted
+    
+    def test_format_message_assistant(self):
+        """Test format_message for assistant role."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, Message
+        
+        tui = PraisonTUI()
+        msg = Message(role="assistant", content="Hi there")
+        
+        formatted = tui._format_message(msg)
+        assert "Hi there" in formatted
+        assert "●" in formatted
+    
+    def test_format_message_tool(self):
+        """Test format_message for tool role."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, Message
+        
+        tui = PraisonTUI()
+        msg = Message(role="tool", content="Running command")
+        
+        formatted = tui._format_message(msg)
+        assert "Running command" in formatted
+        assert "⚙" in formatted
+    
+    def test_handle_command_exit(self):
+        """Test handling exit command."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        tui._running = True
+        
+        result = tui._handle_command("/exit")
+        assert result is True
+        assert tui._running is False
+    
+    def test_handle_command_quit(self):
+        """Test handling quit command."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        tui._running = True
+        
+        result = tui._handle_command("/quit")
+        assert result is True
+        assert tui._running is False
+    
+    def test_handle_command_help(self):
+        """Test handling help command."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        result = tui._handle_command("/help")
+        assert result is True
+        # Help should add a system message
+        assert len(tui.messages) == 1
+        assert tui.messages[0].role == "system"
+    
+    def test_handle_command_clear(self):
+        """Test handling clear command."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI, Message
+        
+        tui = PraisonTUI()
+        tui.messages = [Message(role="user", content="test")]
+        
+        result = tui._handle_command("/clear")
+        assert result is True
+        assert len(tui.messages) == 0
+    
+    def test_handle_command_model_show(self):
+        """Test handling model command (show)."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        result = tui._handle_command("/model")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "gpt-4o-mini" in tui.messages[0].content
+    
+    def test_handle_command_model_change(self):
+        """Test handling model command (change)."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        result = tui._handle_command("/model gpt-4o")
+        assert result is True
+        assert tui.config.model == "gpt-4o"
+    
+    def test_handle_command_cost(self):
+        """Test handling cost command."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        tui._total_tokens = 100
+        tui._total_cost = 0.01
+        
+        result = tui._handle_command("/cost")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "100" in tui.messages[0].content
+    
+    def test_handle_command_compact(self):
+        """Test handling compact command."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        original = tui.config.compact_mode
+        
+        result = tui._handle_command("/compact")
+        assert result is True
+        assert tui.config.compact_mode != original
+    
+    def test_handle_command_multiline(self):
+        """Test handling multiline command."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        original = tui.config.multiline
+        
+        result = tui._handle_command("/multiline")
+        assert result is True
+        assert tui.config.multiline != original
+    
+    def test_handle_command_unknown(self):
+        """Test handling unknown command."""
+        from praisonai.cli.interactive.tui_app import PraisonTUI
+        
+        tui = PraisonTUI()
+        result = tui._handle_command("/unknown")
+        assert result is True
+        assert len(tui.messages) == 1
+        assert "Unknown" in tui.messages[0].content
+
+
+class TestStartTUI:
+    """Tests for start_tui function."""
+    
+    def test_function_exists(self):
+        """Test that start_tui function exists."""
+        from praisonai.cli.interactive.tui_app import start_tui
+        assert callable(start_tui)
+
+
+class TestASCIILogos:
+    """Tests for ASCII logo constants."""
+    
+    def test_logo_large_exists(self):
+        """Test LOGO_LARGE constant exists."""
+        from praisonai.cli.branding import LOGO_LARGE
+        assert len(LOGO_LARGE) > 0
+        assert "PRAISON" in LOGO_LARGE or "██" in LOGO_LARGE
+    
+    def test_logo_small_exists(self):
+        """Test LOGO_SMALL constant exists."""
+        from praisonai.cli.branding import LOGO_SMALL
+        assert len(LOGO_SMALL) > 0
+    
+    def test_logo_tiny_exists(self):
+        """Test LOGO_MEDIUM constant exists (was LOGO_TINY)."""
+        from praisonai.cli.branding import LOGO_MEDIUM
+        assert len(LOGO_MEDIUM) > 0
+    
+    def test_logo_minimal_exists(self):
+        """Test LOGO_MINIMAL constant exists."""
+        from praisonai.cli.branding import LOGO_MINIMAL
+        assert "Praison AI" in LOGO_MINIMAL

src/praisonai/uv.lock

@@ -4276,7 +4276,7 @@ requires-dist = [
     { name = "praisonai-tools", marker = "extra == 'autogen'", specifier = ">=0.1.0" },
     { name = "praisonai-tools", marker = "extra == 'autogen-v4'", specifier = ">=0.1.0" },
     { name = "praisonai-tools", marker = "extra == 'crewai'", specifier = ">=0.1.0" },
-    { name = "praisonaiagents", specifier = ">=0.11.2" },
+    { name = "praisonaiagents", specifier = ">=0.11.6" },
     { name = "pyautogen", marker = "extra == 'autogen'", specifier = "==0.2.29" },
     { name = "pydantic", marker = "extra == 'chat'", specifier = "<=2.10.1" },
     { name = "pydantic", marker = "extra == 'code'", specifier = "<=2.10.1" },
@@ -4324,7 +4324,7 @@ wheels = [
 
 [[package]]
 name = "praisonaiagents"
-version = "0.11.2"
+version = "0.11.6"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "aiohttp" },
@@ -4333,9 +4333,9 @@ dependencies = [
     { name = "pydantic" },
     { name = "rich" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/a4/d0/84583da326312957eb887e32d6df4549889494b736b2f3925f3b80fc7e70/praisonaiagents-0.11.2.tar.gz", hash = "sha256:a32e1063404d61cac71be4ee447f6888f7b9566a298ac63793aeaec41eeeeab1", size = 972927, upload-time = "2026-01-10T19:35:37.032Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/e8/c1/a113cd884d9ca0747f31dc652740b5a859fdc301017f45d75382a0ceba4e/praisonaiagents-0.11.6.tar.gz", hash = "sha256:fb19b18cea0e2cccd7221e34b309dec43d6960463f39de922cd749670d1cae4f", size = 981805, upload-time = "2026-01-12T12:43:47.183Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/43/e8/3c13e6f5a3bcd4b01c0a373a2d3fdd797eeea3f46f673fc6a14ba081e1f2/praisonaiagents-0.11.2-py3-none-any.whl", hash = "sha256:e5d666d23d59e53b498beefc30edef3d28ca42fda7b0cd862d1859e59410067c", size = 954207, upload-time = "2026-01-10T19:35:35.034Z" },
+    { url = "https://files.pythonhosted.org/packages/3d/b5/4129fd74320c367a3d6365fb4481da3a3f7117de36dbb38ea9215e760d1f/praisonaiagents-0.11.6-py3-none-any.whl", hash = "sha256:768288047146ba9431dc23c07b222ebb97d1e05c648f6f3a37cfa71d59a30cdb", size = 964812, upload-time = "2026-01-12T12:43:44.786Z" },
 ]
 
 [[package]]