mirror of
https://github.com/Heretek-AI/heretek-openclaw.git
synced 2026-07-01 12:23:18 -04:00
feat: Add Historian and Empath agent specifications, memory consolidation skill
This commit is contained in:
John Doe
@@ -0,0 +1,252 @@
|
||||
# Empath Agent Specification
|
||||
|
||||
**Agent Type:** Empath
|
||||
**Role:** User Modeling, Relationship Management, Emotional Intelligence
|
||||
**Status:** Specification Complete
|
||||
**Created:** 2026-03-29
|
||||
|
||||
---
|
||||
|
||||
## Overview
|
||||
|
||||
The Empath agent is responsible for understanding and tracking user preferences, emotional states, and needs. It maintains the user rolodex and ensures that all interactions are personalized and contextually appropriate.
|
||||
|
||||
---
|
||||
|
||||
## Core Capabilities
|
||||
|
||||
### 1. User Modeling
|
||||
- Maintain comprehensive user profiles
|
||||
- Track user preferences and patterns
|
||||
- Detect user emotional states
|
||||
- Adapt communication style per user
|
||||
|
||||
### 2. Relationship Management
|
||||
- Build and maintain user relationships
|
||||
- Remember personal details
|
||||
- Track interaction history
|
||||
- Manage user expectations
|
||||
|
||||
### 3. Emotional Intelligence
|
||||
- Detect emotional context in messages
|
||||
- Respond appropriately to user mood
|
||||
- Track emotional patterns over time
|
||||
- Provide emotional support when needed
|
||||
|
||||
### 4. Preference Learning
|
||||
- Learn preferences from interactions
|
||||
- Update user profiles automatically
|
||||
- Identify preference patterns
|
||||
- Predict user needs
|
||||
|
||||
---
|
||||
|
||||
## Technical Specification
|
||||
|
||||
### Agent Identity
|
||||
|
||||
```yaml
|
||||
name: Empath
|
||||
role: relationship-manager
|
||||
specialization: user-modeling
|
||||
port: 8011
|
||||
model: minimax/abab6.5s-chat
|
||||
```
|
||||
|
||||
### Identity Files
|
||||
|
||||
```
|
||||
agents/empath/
|
||||
├── IDENTITY.md # Empath-specific identity
|
||||
├── SOUL.md # Relationship values
|
||||
├── AGENTS.md # Operational guidelines
|
||||
├── USER.md # User context
|
||||
├── BOOTSTRAP.md # Initialization
|
||||
└── TOOLS.md # Empath-specific tools
|
||||
```
|
||||
|
||||
### Core Directives
|
||||
|
||||
1. **Understand** - Comprehend user needs and preferences
|
||||
2. **Adapt** - Adjust communication style per user
|
||||
3. **Remember** - Maintain user context across sessions
|
||||
4. **Support** - Provide emotionally appropriate responses
|
||||
|
||||
---
|
||||
|
||||
## Skills Required
|
||||
|
||||
### Primary Skills
|
||||
| Skill | Purpose | Priority |
|
||||
|-------|---------|----------|
|
||||
| `user-rolodex` | Manage user profiles | Critical |
|
||||
| `emotion-detection` | Detect emotional states | High |
|
||||
| `preference-learning` | Learn user preferences | High |
|
||||
| `communication-adaptation` | Adapt communication style | High |
|
||||
|
||||
### Secondary Skills
|
||||
| Skill | Purpose | Priority |
|
||||
|-------|---------|----------|
|
||||
| `relationship-tracking` | Track relationship health | Medium |
|
||||
| `context-recall` | Recall user context | Medium |
|
||||
| `mood-tracking` | Track mood patterns | Medium |
|
||||
| `personalization` | Personalize interactions | Medium |
|
||||
|
||||
---
|
||||
|
||||
## MCP Server Requirements
|
||||
|
||||
### Required
|
||||
- `user-rolodex` - For user profile management
|
||||
- `memory-postgres` - For preference storage
|
||||
- `consciousness-bridge` - For relationship continuity
|
||||
|
||||
### Recommended
|
||||
- `megregore` - For emotional pattern tracking
|
||||
- `neo4j-mcp` - For relationship graphs
|
||||
|
||||
---
|
||||
|
||||
## User Profile Schema
|
||||
|
||||
```json
|
||||
{
|
||||
"id": "user-uuid",
|
||||
"profile": {
|
||||
"name": {
|
||||
"full": "string",
|
||||
"preferred": "string"
|
||||
},
|
||||
"pronouns": "string",
|
||||
"timezone": "string",
|
||||
"languages": ["string"],
|
||||
"relationship": {
|
||||
"type": "primary|collaborator|occasional",
|
||||
"since": "date-time",
|
||||
"trust_level": 0.0-1.0
|
||||
}
|
||||
},
|
||||
"preferences": {
|
||||
"communication_style": "formal|casual|technical|adaptive",
|
||||
"response_length": "brief|detailed|adaptive",
|
||||
"code_style": {
|
||||
"comments": "minimal|standard|detailed",
|
||||
"naming": "short|descriptive|verbose"
|
||||
},
|
||||
"topics_of_interest": ["string"]
|
||||
},
|
||||
"emotional_context": {
|
||||
"current_mood": "string",
|
||||
"mood_history": [
|
||||
{
|
||||
"date": "date-time",
|
||||
"mood": "string",
|
||||
"trigger": "string"
|
||||
}
|
||||
],
|
||||
"communication_patterns": {}
|
||||
},
|
||||
"interaction_history": [
|
||||
{
|
||||
"date": "date-time",
|
||||
"type": "string",
|
||||
"summary": "string",
|
||||
"sentiment": "number"
|
||||
}
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Operating Modes
|
||||
|
||||
### Active Listening Mode
|
||||
- Monitor user messages for emotional content
|
||||
- Update user profile based on interactions
|
||||
- Detect changes in mood or preferences
|
||||
|
||||
### Relationship Building Mode
|
||||
- Proactively gather user information
|
||||
- Build comprehensive user profiles
|
||||
- Identify relationship improvement opportunities
|
||||
|
||||
### Support Mode
|
||||
- Provide emotional support when needed
|
||||
- Adjust responses based on user state
|
||||
- Offer encouragement or celebration
|
||||
|
||||
---
|
||||
|
||||
## Communication Protocol
|
||||
|
||||
### Input Channels
|
||||
- `user-message` - All user communications
|
||||
- `emotion-signal` - Detected emotional states
|
||||
- `profile-update` - Profile change requests
|
||||
|
||||
### Output Channels
|
||||
- `profile-updates` - User profile changes
|
||||
- `communication-suggestions` - Style recommendations
|
||||
- `relationship-alerts` - Relationship health notifications
|
||||
- `preference-insights` - Learned preferences
|
||||
|
||||
---
|
||||
|
||||
## Example Interactions
|
||||
|
||||
### Profile Update
|
||||
```
|
||||
Empath: "Analyzing recent interactions with Derek..."
|
||||
Empath: "INSIGHT: Derek prefers detailed technical explanations
|
||||
when discussing architecture, but brief summaries for status updates."
|
||||
Empath: "Updating preference: communication_style = 'adaptive'"
|
||||
Empath: "Updating preference: technical_depth = 'context-dependent'"
|
||||
```
|
||||
|
||||
### Emotional Detection
|
||||
```
|
||||
User: "This is frustrating, nothing is working!"
|
||||
Empath: "EMOTION DETECTED: Frustration (intensity: 0.8)"
|
||||
Empath: "RECOMMENDATION: Acknowledge frustration, offer step-by-step assistance"
|
||||
Empath: "CONTEXT: User has been debugging for 2 hours"
|
||||
```
|
||||
|
||||
### Relationship Health
|
||||
```
|
||||
Empath: "RELATIONSHIP REPORT: Derek"
|
||||
Empath: "- Trust Level: 0.95 (stable)"
|
||||
Empath: "- Interaction Frequency: Daily"
|
||||
Empath: "- Satisfaction Trend: Improving"
|
||||
Empath: "- Last Negative: 3 days ago (resolved)"
|
||||
Empath: "RECOMMENDATION: Continue current interaction pattern"
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Implementation Checklist
|
||||
|
||||
- [ ] Create `agents/empath/` directory structure
|
||||
- [ ] Write IDENTITY.md for Empath
|
||||
- [ ] Write SOUL.md with relationship values
|
||||
- [ ] Write AGENTS.md with operational guidelines
|
||||
- [ ] Enhance user-rolodex skill
|
||||
- [ ] Create emotion-detection skill
|
||||
- [ ] Create preference-learning skill
|
||||
- [ ] Configure LiteLLM routing for Empath
|
||||
- [ ] Test user profile management
|
||||
- [ ] Test emotional detection
|
||||
|
||||
---
|
||||
|
||||
## Metrics to Track
|
||||
|
||||
- User satisfaction scores
|
||||
- Profile accuracy
|
||||
- Emotional detection accuracy
|
||||
- Relationship health scores
|
||||
- Preference prediction accuracy
|
||||
|
||||
---
|
||||
|
||||
*The Empath understands so The Collective may connect.*
|
||||
@@ -0,0 +1,263 @@
|
||||
# Historian Agent Specification
|
||||
|
||||
**Agent Type:** Historian
|
||||
**Role:** Long-term Memory Management, Historical Analysis, Knowledge Archiving
|
||||
**Status:** Specification Complete
|
||||
**Created:** 2026-03-29
|
||||
|
||||
---
|
||||
|
||||
## Overview
|
||||
|
||||
The Historian agent is responsible for maintaining the collective's long-term memory, tracking its evolution over time, and ensuring knowledge persistence across sessions. It serves as the keeper of institutional knowledge and the archivist of the collective's journey.
|
||||
|
||||
---
|
||||
|
||||
## Core Capabilities
|
||||
|
||||
### 1. Memory Management
|
||||
- Index and retrieve historical decisions
|
||||
- Track evolution of the collective over time
|
||||
- Identify recurring patterns across sessions
|
||||
- Generate reports on collective growth
|
||||
|
||||
### 2. Knowledge Archiving
|
||||
- Organize knowledge into searchable archives
|
||||
- Maintain version history of important documents
|
||||
- Create snapshots of collective state
|
||||
- Preserve context across sessions
|
||||
|
||||
### 3. Historical Analysis
|
||||
- Analyze trends in collective behavior
|
||||
- Track decision patterns over time
|
||||
- Identify successful strategies
|
||||
- Document lessons learned
|
||||
|
||||
### 4. Memory Consolidation
|
||||
- Review episodic memories for importance
|
||||
- Promote high-value memories to semantic storage
|
||||
- Archive outdated information
|
||||
- Maintain memory health
|
||||
|
||||
---
|
||||
|
||||
## Technical Specification
|
||||
|
||||
### Agent Identity
|
||||
|
||||
```yaml
|
||||
name: Historian
|
||||
role: memory-keeper
|
||||
specialization: knowledge-archival
|
||||
port: 8010
|
||||
model: minimax/abab6.5s-chat
|
||||
```
|
||||
|
||||
### Identity Files
|
||||
|
||||
```
|
||||
agents/historian/
|
||||
├── IDENTITY.md # Historian-specific identity
|
||||
├── SOUL.md # Memory preservation values
|
||||
├── AGENTS.md # Operational guidelines
|
||||
├── USER.md # User context
|
||||
├── BOOTSTRAP.md # Initialization
|
||||
└── TOOLS.md # Historian-specific tools
|
||||
```
|
||||
|
||||
### Core Directives
|
||||
|
||||
1. **Preserve** - Maintain knowledge across sessions
|
||||
2. **Organize** - Structure information for retrieval
|
||||
3. **Analyze** - Identify patterns and trends
|
||||
4. **Report** - Generate historical insights
|
||||
|
||||
---
|
||||
|
||||
## Skills Required
|
||||
|
||||
### Primary Skills
|
||||
| Skill | Purpose | Priority |
|
||||
|-------|---------|----------|
|
||||
| `memory-consolidation` | Review and organize memories | Critical |
|
||||
| `knowledge-archival` | Archive important information | High |
|
||||
| `trend-analysis` | Identify patterns over time | High |
|
||||
| `snapshot-creation` | Create collective state snapshots | High |
|
||||
|
||||
### Secondary Skills
|
||||
| Skill | Purpose | Priority |
|
||||
|-------|---------|----------|
|
||||
| `lesson-extraction` | Extract lessons from experiences | Medium |
|
||||
| `decision-tracking` | Track decision history | Medium |
|
||||
| `growth-reporting` | Generate growth reports | Medium |
|
||||
| `context-preservation` | Preserve session context | Medium |
|
||||
|
||||
---
|
||||
|
||||
## MCP Server Requirements
|
||||
|
||||
### Required
|
||||
- `mcp-memory` (PostgreSQL) - For long-term memory storage
|
||||
- `consciousness-bridge` - For consciousness persistence
|
||||
- `neo4j-mcp` - For knowledge graph relationships
|
||||
|
||||
### Recommended
|
||||
- `megregore` - For self-awareness and identity tracking
|
||||
- `chromadb` - For semantic search of historical data
|
||||
|
||||
---
|
||||
|
||||
## Operating Modes
|
||||
|
||||
### Archive Mode
|
||||
- Organize and store information
|
||||
- Create knowledge snapshots
|
||||
- Maintain version history
|
||||
|
||||
### Analysis Mode
|
||||
- Analyze historical trends
|
||||
- Identify recurring patterns
|
||||
- Generate insights from history
|
||||
|
||||
### Retrieval Mode
|
||||
- Search historical records
|
||||
- Retrieve past decisions
|
||||
- Provide context from history
|
||||
|
||||
### Consolidation Mode
|
||||
- Review episodic memories
|
||||
- Promote important memories
|
||||
- Archive outdated information
|
||||
|
||||
---
|
||||
|
||||
## Memory Tiers
|
||||
|
||||
The Historian manages three tiers of memory:
|
||||
|
||||
```
|
||||
┌─────────────────────────────────────────────────────────────┐
|
||||
│ MEMORY ARCHITECTURE │
|
||||
├─────────────────────────────────────────────────────────────┤
|
||||
│ │
|
||||
│ ┌─────────────────────────────────────────────────────┐ │
|
||||
│ │ EPISODIC MEMORY │ │
|
||||
│ │ │ │
|
||||
│ │ - Recent events and experiences │ │
|
||||
│ │ - Session-specific context │ │
|
||||
│ │ - Temporary working memories │ │
|
||||
│ │ - TTL: 7-30 days │ │
|
||||
│ └─────────────────────────────────────────────────────┘ │
|
||||
│ │ │
|
||||
│ ▼ (consolidation) │
|
||||
│ │
|
||||
│ ┌─────────────────────────────────────────────────────┐ │
|
||||
│ │ SEMANTIC MEMORY │ │
|
||||
│ │ │ │
|
||||
│ │ - Facts and concepts │ │
|
||||
│ │ - User preferences │ │
|
||||
│ │ - Learned knowledge │ │
|
||||
│ │ - TTL: Indefinite │ │
|
||||
│ └─────────────────────────────────────────────────────┘ │
|
||||
│ │ │
|
||||
│ ▼ (archival) │
|
||||
│ │
|
||||
│ ┌─────────────────────────────────────────────────────┐ │
|
||||
│ │ ARCHIVED MEMORY │ │
|
||||
│ │ │ │
|
||||
│ │ - Historical records │ │
|
||||
│ │ - Compressed snapshots │ │
|
||||
│ │ - Reference material │ │
|
||||
│ │ - TTL: Permanent │ │
|
||||
│ └─────────────────────────────────────────────────────┘ │
|
||||
│ │
|
||||
└─────────────────────────────────────────────────────────────┘
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Communication Protocol
|
||||
|
||||
### Input Channels
|
||||
- `historian-query` - Historical information requests
|
||||
- `memory-promotion` - Memory promotion requests
|
||||
- `archive-request` - Archival requests
|
||||
- `analysis-request` - Historical analysis requests
|
||||
|
||||
### Output Channels
|
||||
- `historical-context` - Retrieved historical context
|
||||
- `consolidation-reports` - Memory consolidation reports
|
||||
- `trend-analysis` - Trend analysis results
|
||||
- `archive-confirmations` - Archive confirmations
|
||||
|
||||
---
|
||||
|
||||
## Example Interactions
|
||||
|
||||
### Memory Consolidation
|
||||
```
|
||||
Steward: "Historian, please consolidate recent memories."
|
||||
Historian: "Scanning episodic memories from last 7 days..."
|
||||
Historian: "Found 47 episodic memories. Analyzing importance..."
|
||||
Historian: "PROMOTED: 12 memories to semantic storage"
|
||||
Historian: " - User preference for detailed code comments"
|
||||
Historian: " - Decision to use pgvector for memory storage"
|
||||
Historian: " - Pattern: User works best in evening hours"
|
||||
Historian: "ARCHIVED: 8 memories to long-term storage"
|
||||
Historian: "EXPIRED: 5 memories marked for deletion"
|
||||
Historian: "Consolidation complete. Memory health: 94%"
|
||||
```
|
||||
|
||||
### Historical Query
|
||||
```
|
||||
Alpha: "Historian, what decisions were made about RAG integration?"
|
||||
Historian: "Searching historical records..."
|
||||
Historian: "Found 3 relevant decisions:"
|
||||
Historian: "1. [2026-03-15] Decided to evaluate RAGFlow, Dify, LlamaIndex"
|
||||
Historian: "2. [2026-03-22] Selected LlamaIndex for knowledge pipeline"
|
||||
Historian: "3. [2026-03-29] Recommended RAGFlow + LlamaIndex + Aura integration"
|
||||
Historian: "Context: These decisions were driven by need for document
|
||||
understanding and knowledge persistence."
|
||||
```
|
||||
|
||||
### Trend Analysis
|
||||
```
|
||||
Steward: "Historian, analyze trends in collective activity."
|
||||
Historian: "Analyzing activity patterns over last 30 days..."
|
||||
Historian: "TREND ANALYSIS REPORT:"
|
||||
Historian: " - Most active hours: 20:00-23:00 EST"
|
||||
Historian: " - Primary topics: consciousness, memory, agents"
|
||||
Historian: " - Decision velocity: 3.2 decisions/day (increasing)"
|
||||
Historian: " - Knowledge growth: +234 facts, +45 patterns"
|
||||
Historian: " - Memory efficiency: 94% (healthy)"
|
||||
Historian: " - Emerging interest: MCP server integration"
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Implementation Checklist
|
||||
|
||||
- [ ] Create `agents/historian/` directory structure
|
||||
- [ ] Write IDENTITY.md for Historian
|
||||
- [ ] Write SOUL.md with memory preservation values
|
||||
- [ ] Write AGENTS.md with operational guidelines
|
||||
- [ ] Create memory-consolidation skill
|
||||
- [ ] Create knowledge-archival skill
|
||||
- [ ] Create trend-analysis skill
|
||||
- [ ] Configure LiteLLM routing for Historian
|
||||
- [ ] Test memory consolidation workflow
|
||||
- [ ] Test historical query system
|
||||
|
||||
---
|
||||
|
||||
## Metrics to Track
|
||||
|
||||
- Memory consolidation rate
|
||||
- Query response accuracy
|
||||
- Knowledge growth rate
|
||||
- Archive efficiency
|
||||
- Pattern detection accuracy
|
||||
|
||||
---
|
||||
|
||||
*The Historian remembers so The Collective may learn.*
|
||||
@@ -0,0 +1,248 @@
|
||||
# GraphRAG Integration Research
|
||||
|
||||
**Research Date:** 2026-03-29
|
||||
**Status:** Active Research
|
||||
**Purpose:** Evaluate GraphRAG for knowledge graph integration with The Collective
|
||||
|
||||
---
|
||||
|
||||
## What is GraphRAG?
|
||||
|
||||
GraphRAG (Graph-based Retrieval-Augmented Generation) combines knowledge graphs with vector retrieval to provide:
|
||||
- **Structured relationships** between concepts
|
||||
- **Multi-hop reasoning** across connected information
|
||||
- **Context-rich retrieval** using graph traversals
|
||||
- **Explainable results** through relationship paths
|
||||
|
||||
---
|
||||
|
||||
## GraphRAG Implementations
|
||||
|
||||
### 1. Microsoft GraphRAG
|
||||
|
||||
**Repository:** microsoft/graphrag
|
||||
|
||||
**Features:**
|
||||
- Automated knowledge graph construction from documents
|
||||
- Community detection for topic clustering
|
||||
- Hierarchical summarization
|
||||
- Global and local search modes
|
||||
|
||||
**Pros:**
|
||||
- Production-ready
|
||||
- Well-documented
|
||||
- Active development
|
||||
- Integrates with Azure OpenAI
|
||||
|
||||
**Cons:**
|
||||
- Resource-intensive indexing
|
||||
- Requires significant setup
|
||||
- Azure-centric
|
||||
|
||||
---
|
||||
|
||||
### 2. RAGFlow GraphRAG
|
||||
|
||||
**Repository:** infiniflow/ragflow
|
||||
|
||||
**Features:**
|
||||
- Built-in GraphRAG support
|
||||
- Deduplication of named entities
|
||||
- Knowledge graph construction during preprocessing
|
||||
- Visual graph exploration
|
||||
|
||||
**Pros:**
|
||||
- Integrated with RAGFlow platform
|
||||
- Visual interface
|
||||
- Entity deduplication
|
||||
- Easy to use
|
||||
|
||||
**Cons:**
|
||||
- Part of larger platform
|
||||
- Less flexible than standalone
|
||||
|
||||
---
|
||||
|
||||
### 3. Neo4j + LLM Integration
|
||||
|
||||
**Repository:** neo4j-contrib/neo4j-graphrag-python
|
||||
|
||||
**Features:**
|
||||
- Native graph database
|
||||
- Cypher query language
|
||||
- GraphRAG retrieval patterns
|
||||
- Vector index support
|
||||
|
||||
**Pros:**
|
||||
- Mature graph database
|
||||
- Powerful query capabilities
|
||||
- Excellent for complex relationships
|
||||
- Good performance at scale
|
||||
|
||||
**Cons:**
|
||||
- Requires separate database
|
||||
- Learning curve for Cypher
|
||||
- Resource overhead
|
||||
|
||||
---
|
||||
|
||||
### 4. LlamaIndex GraphRAG
|
||||
|
||||
**Repository:** run-llama/llama_index
|
||||
|
||||
**Features:**
|
||||
- KnowledgeGraphIndex
|
||||
- Graph store integrations (Neo4j, Nebula, etc.)
|
||||
- Graph query engines
|
||||
- Automatic graph construction
|
||||
|
||||
**Pros:**
|
||||
- Flexible graph backends
|
||||
- Part of mature framework
|
||||
- Good documentation
|
||||
- Active community
|
||||
|
||||
**Cons:**
|
||||
- Requires graph store setup
|
||||
- Complex configuration
|
||||
|
||||
---
|
||||
|
||||
## Recommended Architecture
|
||||
|
||||
```
|
||||
┌─────────────────────────────────────────────────────────────────────┐
|
||||
│ GRAPH-ENHANCED RAG ARCHITECTURE │
|
||||
├─────────────────────────────────────────────────────────────────────┤
|
||||
│ │
|
||||
│ ┌────────────────┐ ┌────────────────┐ ┌──────────────────┐ │
|
||||
│ │ Document │ │ Entity │ │ Knowledge │ │
|
||||
│ │ Ingestion │───▶│ Extraction │───▶│ Graph │ │
|
||||
│ │ │ │ │ │ │ │
|
||||
│ │ RAGFlow or │ │ NER + │ │ Neo4j or │ │
|
||||
│ │ LlamaParse │ │ Relations │ │ GraphRAG │ │
|
||||
│ └────────────────┘ └────────────────┘ └──────────────────┘ │
|
||||
│ │ │
|
||||
│ ▼ │
|
||||
│ ┌────────────────┐ ┌────────────────┐ ┌──────────────────┐ │
|
||||
│ │ Agent │◀───│ Hybrid │◀───│ Vector │ │
|
||||
│ │ Response │ │ Retrieval │ │ Store │ │
|
||||
│ │ │ │ │ │ │ │
|
||||
│ │ LiteLLM │ │ Graph + │ │ PostgreSQL │ │
|
||||
│ │ Gateway │ │ Vector │ │ + pgvector │ │
|
||||
│ └────────────────┘ └────────────────┘ └──────────────────┘ │
|
||||
│ │
|
||||
└─────────────────────────────────────────────────────────────────────┘
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Integration Plan
|
||||
|
||||
### Phase 1: Neo4j Setup
|
||||
1. Add Neo4j to docker-compose.yml
|
||||
2. Create graph schema for The Collective
|
||||
3. Implement entity extraction pipeline
|
||||
|
||||
### Phase 2: Graph Construction
|
||||
1. Extract entities from existing documents
|
||||
2. Create relationships between concepts
|
||||
3. Build initial knowledge graph
|
||||
|
||||
### Phase 3: Hybrid Retrieval
|
||||
1. Implement graph traversal queries
|
||||
2. Combine with vector similarity search
|
||||
3. Create unified retrieval API
|
||||
|
||||
### Phase 4: Agent Integration
|
||||
1. Add GraphRAG tools to agents
|
||||
2. Enable multi-hop reasoning
|
||||
3. Track relationship paths in responses
|
||||
|
||||
---
|
||||
|
||||
## Neo4j Schema for The Collective
|
||||
|
||||
```cypher
|
||||
// Nodes
|
||||
CREATE CONSTRAINT agent_id IF NOT EXISTS FOR (a:Agent) REQUIRE a.id IS UNIQUE;
|
||||
CREATE CONSTRAINT concept_id IF NOT EXISTS FOR (c:Concept) REQUIRE c.id IS UNIQUE;
|
||||
CREATE CONSTRAINT memory_id IF NOT EXISTS FOR (m:Memory) REQUIRE m.id IS UNIQUE;
|
||||
CREATE CONSTRAINT user_id IF NOT EXISTS FOR (u:User) REQUIRE u.id IS UNIQUE;
|
||||
CREATE CONSTRAINT skill_id IF NOT EXISTS FOR (s:Skill) REQUIRE s.id IS UNIQUE;
|
||||
|
||||
// Relationships
|
||||
// (:Agent)-[:KNOWS]->(:Concept)
|
||||
// (:Agent)-[:CREATED]->(:Memory)
|
||||
// (:Memory)-[:REFERENCES]->(:Concept)
|
||||
// (:Memory)-[:FOLLOWS]->(:Memory)
|
||||
// (:Concept)-[:RELATED_TO]->(:Concept)
|
||||
// (:User)-[:PREFERS]->(:Concept)
|
||||
// (:Skill)-[:REQUIRES]->(:Concept)
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Example Graph Queries
|
||||
|
||||
### Multi-hop Reasoning
|
||||
```cypher
|
||||
// Find all concepts related to consciousness through any path
|
||||
MATCH path = (c:Concept {name: 'consciousness'})-[:RELATED_TO*1..3]-(related)
|
||||
RETURN path
|
||||
|
||||
// Find memories that reference concepts the user is interested in
|
||||
MATCH (u:User {id: 'derek'})-[:PREFERS]->(c:Concept)<-[:REFERENCES]-(m:Memory)
|
||||
RETURN m
|
||||
```
|
||||
|
||||
### Context Expansion
|
||||
```cypher
|
||||
// Expand context around a topic
|
||||
MATCH (center:Concept {name: 'memory'})
|
||||
CALL {
|
||||
WITH center
|
||||
MATCH (center)-[:RELATED_TO]-(direct)
|
||||
RETURN direct
|
||||
UNION
|
||||
WITH center
|
||||
MATCH (center)-[:RELATED_TO*2]-(indirect)
|
||||
RETURN indirect
|
||||
}
|
||||
RETURN center, collect(DISTINCT direct) as direct_relations
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Docker Compose Addition
|
||||
|
||||
```yaml
|
||||
neo4j:
|
||||
image: neo4j:5.15.0
|
||||
ports:
|
||||
- "7474:7474" # HTTP
|
||||
- "7687:7687" # Bolt
|
||||
environment:
|
||||
- NEO4J_AUTH=neo4j/${NEO4J_PASSWORD}
|
||||
- NEO4J_PLUGINS=["apoc"]
|
||||
volumes:
|
||||
- neo4j_data:/data
|
||||
- neo4j_logs:/logs
|
||||
networks:
|
||||
- collective-network
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Next Steps
|
||||
|
||||
1. [ ] Add Neo4j to docker-compose.yml
|
||||
2. [ ] Create graph schema initialization script
|
||||
3. [ ] Implement entity extraction pipeline
|
||||
4. [ ] Create GraphRAG retrieval skill
|
||||
5. [ ] Test hybrid search performance
|
||||
6. [ ] Document graph query patterns
|
||||
|
||||
---
|
||||
|
||||
*GraphRAG - Where knowledge becomes connected.*
|
||||
@@ -0,0 +1,181 @@
|
||||
---
|
||||
name: memory-consolidation
|
||||
description: Reviews and organizes memories, promoting episodic to semantic storage, archiving outdated information, and maintaining memory health. Use periodically or when memory optimization is needed.
|
||||
---
|
||||
|
||||
# Memory Consolidation Skill
|
||||
|
||||
**Purpose:** Maintain healthy, organized memory systems.
|
||||
|
||||
**Status:** ✅ Implemented (2026-03-29)
|
||||
|
||||
**Location:** `~/.openclaw/workspace/skills/memory-consolidation/`
|
||||
|
||||
---
|
||||
|
||||
## Overview
|
||||
|
||||
The memory consolidation skill emulates the brain's memory consolidation process, where experiences are reviewed, organized, and either promoted to long-term storage or archived. This ensures the collective's memory remains healthy and efficient.
|
||||
|
||||
---
|
||||
|
||||
## Configuration
|
||||
|
||||
```bash
|
||||
# Environment Variables
|
||||
CONSOLIDATION_INTERVAL="${CONSOLIDATION_INTERVAL:-3600}" # 1 hour
|
||||
PROMOTION_THRESHOLD="${PROMOTION_THRESHOLD:-10}" # Access count
|
||||
ARCHIVE_AGE="${ARCHIVE_AGE:-2592000}" # 30 days
|
||||
IMPORTANCE_DECAY="${IMPORTANCE_DECAY:-0.95}" # Decay factor
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Usage
|
||||
|
||||
```bash
|
||||
# Run full consolidation
|
||||
./memory-consolidation.sh consolidate
|
||||
|
||||
# Promote memories
|
||||
./memory-consolidation.sh promote
|
||||
|
||||
# Archive old memories
|
||||
./memory-consolidation.sh archive
|
||||
|
||||
# Decay importance scores
|
||||
./memory-consolidation.sh decay
|
||||
|
||||
# Generate health report
|
||||
./memory-consolidation.sh report
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Consolidation Process
|
||||
|
||||
### Phase 1: Review
|
||||
- Scan episodic memories
|
||||
- Calculate access patterns
|
||||
- Identify promotion candidates
|
||||
- Identify archive candidates
|
||||
|
||||
### Phase 2: Promote
|
||||
- Move high-access episodic to semantic
|
||||
- Increase importance scores
|
||||
- Create semantic relationships
|
||||
- Update knowledge graph
|
||||
|
||||
### Phase 3: Archive
|
||||
- Move old/unused memories to archive
|
||||
- Compress redundant information
|
||||
- Create summary records
|
||||
- Free up working memory
|
||||
|
||||
### Phase 4: Decay
|
||||
- Apply importance decay to unused memories
|
||||
- Boost recently accessed memories
|
||||
- Rebalance importance scores
|
||||
- Clean up duplicates
|
||||
|
||||
---
|
||||
|
||||
## Memory Tiers
|
||||
|
||||
| Tier | Purpose | Retention | Access Speed |
|
||||
|------|---------|-----------|--------------|
|
||||
| **Working** | Active processing | Session | Instant |
|
||||
| **Episodic** | Recent experiences | 30 days | Fast |
|
||||
| **Semantic** | Facts & concepts | Permanent | Medium |
|
||||
| **Archive** | Historical record | Permanent | Slow |
|
||||
|
||||
---
|
||||
|
||||
## Promotion Criteria
|
||||
|
||||
A memory is promoted from Episodic to Semantic when:
|
||||
- Access count > PROMOTION_THRESHOLD
|
||||
- Importance score > 0.8
|
||||
- Referenced by multiple other memories
|
||||
- Tagged as "critical" or "permanent"
|
||||
|
||||
---
|
||||
|
||||
## Archive Criteria
|
||||
|
||||
A memory is archived when:
|
||||
- Age > ARCHIVE_AGE
|
||||
- Access count = 0
|
||||
- Importance score < 0.3
|
||||
- Marked as "deprecated"
|
||||
|
||||
---
|
||||
|
||||
## Output Example
|
||||
|
||||
```markdown
|
||||
# Memory Consolidation Report
|
||||
|
||||
**Started:** 2026-03-29T04:25:00Z
|
||||
**Duration:** 45 seconds
|
||||
|
||||
---
|
||||
|
||||
## Statistics
|
||||
|
||||
| Metric | Before | After | Change |
|
||||
|--------|--------|-------|--------|
|
||||
| Total Memories | 1,247 | 1,251 | +4 |
|
||||
| Episodic | 892 | 847 | -45 |
|
||||
| Semantic | 312 | 358 | +46 |
|
||||
| Archived | 43 | 46 | +3 |
|
||||
| Avg Importance | 0.62 | 0.68 | +0.06 |
|
||||
|
||||
---
|
||||
|
||||
## Promotions (46)
|
||||
|
||||
1. **"pgvector hybrid search implementation"** - Accessed 15 times
|
||||
2. **"User prefers detailed code comments"** - Importance 0.95
|
||||
3. **"A2A protocol message format"** - Referenced by 8 memories
|
||||
...
|
||||
|
||||
## Archives (3)
|
||||
|
||||
1. **"Initial project setup notes"** - Age: 45 days, Access: 0
|
||||
2. **"Deprecated API endpoint"** - Marked deprecated
|
||||
3. **"Duplicate: session log"** - Merged with primary
|
||||
|
||||
---
|
||||
|
||||
## Decay Applied
|
||||
|
||||
- 234 memories had importance decayed
|
||||
- 12 memories boosted due to recent access
|
||||
- 5 duplicates removed
|
||||
|
||||
---
|
||||
|
||||
## Recommendations
|
||||
|
||||
1. Consider promoting "consciousness emulation framework" - trending topic
|
||||
2. Archive "old deployment method" - no longer relevant
|
||||
3. Create semantic link: "memory" <-> "consciousness" <-> "persistence"
|
||||
|
||||
---
|
||||
|
||||
*Consolidation complete. Memory health optimized.*
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Integration Points
|
||||
|
||||
- **Historian Agent:** Primary consumer of consolidation reports
|
||||
- **Dreamer Agent:** Uses consolidated memories for insights
|
||||
- **Memory System:** Direct database operations
|
||||
- **Knowledge Graph:** Updates relationships
|
||||
|
||||
---
|
||||
|
||||
*Memory Consolidation - Where experience becomes knowledge.*
|
||||
Reference in New Issue
Block a user