Willow 🌳

Autonomous Agent Population & Reasoning System

Current Status: Bootstrapping Phase Mission Control: See MISSION_CONTROL.md for links, status, and objectives.

Overview

Willow is a self-describing, introspective system where skills, workflows, brand assets, and conversational memory are stored AS graph nodes in Neo4j. Claude can query "what can I do?" by querying the ontology itself.

Project Management: Jira Board (SCRUM)

Quick Start

1. Check Mission Control: Ensure you understand the current sprint.
2. Bootstrap:

   pip install -r requirements.txt
   python bootstrap/verify_graph.py

3. Run Hello World:

   python core/skills/hello_willow.py

Architecture

• Brain: Neo4j AuraDB (Graph Memory)
• Vault: Xeon Server (Postgres Population DB)
• Muscle: Windows 11 PC (Inference & Generation)
• Architect: Mac Mini (Control Plane)

See docs/ARCHITECTURE_DEFINITIVE.md for details.

┌─────────────────┐
│  Claude Code    │
│  (AntiGravity)  │
└────────┬────────┘
         │
    ┌────┴────┐
    │   MCP   │
    └────┬────┘
         │
┌────────┴──────────────────────────────┐
│         Docker Infrastructure         │
│                                       │
│  ┌─────────┐  ┌──────────┐          │
│  │ Neo4j   │  │ Neo4j    │          │
│  │ Dev     │  │ MCP      │          │
│  │ :7474   │  │ Server   │          │
│  │ :7687   │  │ :3001    │          │
│  └─────────┘  └──────────┘          │
│                                       │
│  ┌─────────┐  ┌──────────┐          │
│  │ Willow  │  │   N8N    │          │
│  │ API     │  │ :5678    │          │
│  │ :8000   │  └──────────┘          │
│  └─────────┘                         │
└───────────────────────────────────────┘

Quick Start

1. Start the infrastructure

docker-compose up -d

2. Load the bootstrap schema

docker exec -i willow-neo4j cypher-shell -u neo4j -p willowdev123 < schemas/willow-bootstrap.cypher

3. Verify system health

• Neo4j Browser: http://localhost:7474
• Willow API: http://localhost:8000
• Neo4j MCP Server: http://localhost:3001
• N8N: http://localhost:5678

Services

Neo4j (Graph Database)

• Port: 7474 (HTTP), 7687 (Bolt)
• Credentials: neo4j / willowdev123
• APOC plugins enabled

Neo4j MCP Server

• Port: 3001
• Exposes tools: run_cypher, get_skills, execute_skill, get_brand_assets
• Logs all queries as :ExecutionLog nodes

Willow API

• Port: 8000
• Executes Python skills dynamically
• Mounts /skills directory

N8N

• Port: 5678
• Workflow orchestration
• Credentials: willow / willowdev123

Skills

hello_world.py

Simple greeting skill to verify system is operational

execute(name="Peter")
# Returns: "Hello, Peter! Willow is alive."

retrieve_conversation_context.py

Query decisions and insights by keyword

execute(keyword="Docker", depth=2, limit=10)
# Returns: Decisions mentioning "Docker" with related context

ingest_mssql_claims.py

Read from MSSQL and create graph nodes (Phase 5)

execute(table="Claims", limit=100)
# Placeholder - waiting for MSSQL credentials

Graph Schema

Core Node Types

• :System - System metadata
• :Skill - Executable capabilities (Python or Cypher)
• :BrandAsset - Seasonal UI themes and branding
• :Workflow - N8N orchestration definitions
• :Decision - Conversational memory (why we built things)
• :Insight - Key learnings
• :ConversationSession - Discussion history
• :ExecutionLog - Query audit trail

Key Relationships

• (:System)-[:HAS_CAPABILITY]->(:Skill)
• (:System)-[:HAS_DECISION]->(:Decision)
• (:System)-[:HAS_INSIGHT]->(:Insight)
• (:Decision)-[:LED_TO_INSIGHT]->(:Insight)
• (:Skill)-[:IMPLEMENTS_INSIGHT]->(:Insight)

Testing

Verify Neo4j Connection

docker exec -it willow-neo4j cypher-shell -u neo4j -p willowdev123

Query Skills

MATCH (s:Skill) RETURN s.name, s.language, s.description;

Get Active Brand Palette

MATCH (b:BrandAsset {active: true}) RETURN b.palette;

Test Hello World Skill

curl -X POST http://localhost:8000/execute \
  -H "Content-Type: application/json" \
  -d '{"skill_name": "hello_world", "parameters": {"name": "Peter"}}'

Brand Identity

Active Season: Autumn 🍂

Palette:

• Primary: #8B9D83 (muted sage green)
• Accent: #C17A5C (burnt amber)
• Grey: #A89F91
• Burgundy: #6B4444
• Background: #F5F1E8 (warm cream)

Logo: Minimalist geometric willow tree (Canva)

Development Principles

1. Muddy the canvas - Start concrete, iterate
2. Zoomy and invisible - No heavy pipelines
3. Make decisions - Explain reasoning, test ideas
4. R&D mindset - Follow plans not clocks

Next Steps

• Connect to Peter's MSSQL database
• Build graph visualization skill
• Create dashboard using brand assets
• Sync to AuraDB for durability
• Add VR visualization integration

License

Proprietary - Semantic Arts Innovations Team R&D

---

Built with: Neo4j, Docker, FastAPI, Python For: Peter (DBA & Ontologist at Semantic Arts) By: Claude Desktop + Claude Code (AntiGravity)