LangGraph
Agent workflow orchestration engine.
Overview
LangGraph provides the execution environment for AI agent graphs.
| Property | Value |
|---|---|
| Container | langgraph |
| URL | norns-api.ravenhelm.dev |
| Port | 2024 (internal) |
| Source | ~/ravenhelm/docs/AI-ML-Platform/norns-agent/agent/ |
Architecture
LangGraph implements a state machine for agent execution:
Input → State → Node → Conditional Edge → Node → Output
↑ │
└──────────────────────────────────┘
(Loop until done)
Agent Graphs
| Graph | Purpose |
|---|---|
assistant | General conversation |
task_manager | Task operations |
home_automation | Smart home control |
Quick Commands
# View logs
docker logs -f langgraph
# Restart
docker restart langgraph
# Check API
curl https://norns-api.ravenhelm.dev/info
LangGraph Studio
Access the visual graph editor:
# Open LangGraph dev UI
open http://localhost:2024
Configuration
Same environment as Norns Agent (shares codebase):
command: ["langgraph", "dev", "--host", "0.0.0.0", "--port", "2024", "--no-browser"]
Troubleshooting
Issue: Graph Execution Hangs
Symptoms: Agent doesn't respond, no error
Diagnosis:
docker logs langgraph | tail -50
# Check Langfuse for incomplete traces
Solutions:
- Check for infinite loops in graph
- Verify tool execution completes
- Review state transitions