langchain-ai/deepagents
1
0
Fork 0
mirror of https://github.com/langchain-ai/deepagents.git synced 2026-07-22 09:45:24 -04:00
Code Issues 169 Packages Projects Releases 16 Wiki Activity

[PR #74] [MERGED] Added Support for Model Instances and Dict #441

New Issue
Closed
opened 2026-02-16 09:16:02 -05:00 by yindo · 0 comments
yindo commented 2026-02-16 09:16:02 -05:00
Owner
Copy Link
Copy Source

📋 Pull Request Information

Original PR: https://github.com/langchain-ai/deepagents/pull/74
Author: @riadzeitounn
Created: 8/25/2025
Status: ✅ Merged
Merged: 8/25/2025
Merged by: @hwchase17

Base: master ← Head: support-model-instance-format

📝 Commits (1)

  • d2c19ad Added Support for Model Instances and dict

📊 Changes

2 files changed (+15 additions, -8 deletions)

View changed files

📝 src/deepagents/graph.py (+1 -0)
📝 src/deepagents/sub_agent.py (+14 -8)

📄 Description

Improve Subagent Model Configuration: Unified Model Field

The Problem

Right now, there's a weird inconsistency between how the main agent and subagents handle models:

  • The main agent uses proper model instances like ChatAnthropic(model_name="claude-sonnet-4", max_tokens=64000)
  • But subagents have to use dictionary configs like {"model": "anthropic", "model_name": "claude-haiku"}

This is pretty annoying because:

  1. It's inconsistent - why should main agents and subagents work differently?
  2. Code duplication - you have to repeat the same model settings everywhere
  3. It's just not intuitive

My Solution

I updated the SubAgent TypedDict to use a unified model field that intelligently accepts both model instances AND dictionary settings.

Before (the old way):

subagents = [
    {
        "name": "researcher",
        "description": "Does research tasks", 
        "prompt": "You are a researcher...",
        "model_settings": {"model": "anthropic", "model_name": "claude-haiku-3"}  # Ugh, separate field
    }
]

After (unified and flexible):

from langchain_anthropic import ChatAnthropic

# Create model instances
fast_model = ChatAnthropic(model_name="claude-haiku-3", max_tokens=4000)

# Use the same 'model' field for both approaches!
subagents = [
    {
        "name": "researcher", 
        "description": "Does research tasks",
        "prompt": "You are a researcher...",
        "model": fast_model  # ✅ Model instance - clean and reusable!
    },
    {
        "name": "writer",
        "description": "Does writing tasks", 
        "prompt": "You are a writer...",
        "model": {"model": "anthropic", "model_name": "claude-sonnet-4"}  # ✅ Dict settings - still works!
    },
    {
        "name": "analyzer",
        "description": "Does analysis tasks",
        "prompt": "You are an analyzer..."
        # ✅ No model field - uses main model
    }
]

What I Changed

  1. Updated SubAgent TypedDict in src/deepagents/sub_agent.py:

    • Replaced model_settings: NotRequired[dict[str, Any]]
    • With model: NotRequired[Union[LanguageModelLike, dict[str, Any]]]

  2. Smart model resolution logic in _create_task_tool:

    • Detects if model is a dictionary → creates model with init_chat_model(**agent_model)
    • Detects if model is an instance → uses directly
    • Falls back to main model if no model specified

  3. Updated documentation to reflect the unified approach

Benefits

  • ✅ Unified API - one model field handles both approaches
  • ✅ Flexible - accepts both model instances and dictionary settings
  • ✅ Backward compatible - existing dictionary code still works
  • ✅ Type safe - proper typing with Union[LanguageModelLike, dict[str, Any]]
  • ✅ Intuitive - same field name, smart value handling
  • ✅ Less duplication - create models once, reuse them

This makes the subagent model configuration much more intuitive and consistent with the rest of the codebase!

Files Changed:

  • src/deepagents/sub_agent.py - Updated TypedDict and model resolution logic
  • src/deepagents/graph.py - Updated documentation

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/langchain-ai/deepagents/pull/74 **Author:** [@riadzeitounn](https://github.com/riadzeitounn) **Created:** 8/25/2025 **Status:** ✅ Merged **Merged:** 8/25/2025 **Merged by:** [@hwchase17](https://github.com/hwchase17) **Base:** `master` ← **Head:** `support-model-instance-format` --- ### 📝 Commits (1) - [`d2c19ad`](https://github.com/langchain-ai/deepagents/commit/d2c19ad4835ec1f9704782bbb70c7b3f8c88e537) Added Support for Model Instances and dict ### 📊 Changes **2 files changed** (+15 additions, -8 deletions) <details> <summary>View changed files</summary> 📝 `src/deepagents/graph.py` (+1 -0) 📝 `src/deepagents/sub_agent.py` (+14 -8) </details> ### 📄 Description # Improve Subagent Model Configuration: Unified Model Field ## The Problem Right now, there's a weird inconsistency between how the main agent and subagents handle models: - The main agent uses proper model instances like `ChatAnthropic(model_name="claude-sonnet-4", max_tokens=64000)` - But subagents have to use dictionary configs like `{"model": "anthropic", "model_name": "claude-haiku"}` This is pretty annoying because: 1. **It's inconsistent** - why should main agents and subagents work differently? 2. **Code duplication** - you have to repeat the same model settings everywhere 3. **It's just not intuitive** ## My Solution I updated the `SubAgent` TypedDict to use a **unified `model` field** that intelligently accepts **both model instances AND dictionary settings**. ### Before (the old way): ```python subagents = [ { "name": "researcher", "description": "Does research tasks", "prompt": "You are a researcher...", "model_settings": {"model": "anthropic", "model_name": "claude-haiku-3"} # Ugh, separate field } ] ``` ### After (unified and flexible): ```python from langchain_anthropic import ChatAnthropic # Create model instances fast_model = ChatAnthropic(model_name="claude-haiku-3", max_tokens=4000) # Use the same 'model' field for both approaches! subagents = [ { "name": "researcher", "description": "Does research tasks", "prompt": "You are a researcher...", "model": fast_model # ✅ Model instance - clean and reusable! }, { "name": "writer", "description": "Does writing tasks", "prompt": "You are a writer...", "model": {"model": "anthropic", "model_name": "claude-sonnet-4"} # ✅ Dict settings - still works! }, { "name": "analyzer", "description": "Does analysis tasks", "prompt": "You are an analyzer..." # ✅ No model field - uses main model } ] ``` ## What I Changed 1. **Updated `SubAgent` TypedDict** in `src/deepagents/sub_agent.py`: - Replaced `model_settings: NotRequired[dict[str, Any]]` - With `model: NotRequired[Union[LanguageModelLike, dict[str, Any]]]` 2. **Smart model resolution logic** in `_create_task_tool`: - Detects if `model` is a dictionary → creates model with `init_chat_model(**agent_model)` - Detects if `model` is an instance → uses directly - Falls back to main model if no `model` specified 3. **Updated documentation** to reflect the unified approach ## Benefits - ✅ **Unified API** - one `model` field handles both approaches - ✅ **Flexible** - accepts both model instances and dictionary settings - ✅ **Backward compatible** - existing dictionary code still works - ✅ **Type safe** - proper typing with `Union[LanguageModelLike, dict[str, Any]]` - ✅ **Intuitive** - same field name, smart value handling - ✅ **Less duplication** - create models once, reuse them This makes the subagent model configuration much more intuitive and consistent with the rest of the codebase! --- **Files Changed:** - `src/deepagents/sub_agent.py` - Updated TypedDict and model resolution logic - `src/deepagents/graph.py` - Updated documentation --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
yindo added the pull-request label 2026-02-16 09:16:02 -05:00
yindo closed this issue 2026-02-16 09:16:02 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
MCP
backends
bug
cli
config
customization
deepagents
documentation
external
feature
fix
good first issue
help wanted
integration
internal
invalid
middleware
model
needs investigation
p0
p1
p2
p3
performance
pull-request
Mirrored from GitHub Pull Request
 question
sdk
subagents
unable to reproduce
windows
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langchain-ai/deepagents#441
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?