langgraph-dev-navigator

LangGraph-Dev-Navigator is a grounded AI assistant designed to operate against a specific, version-controlled LangGraph codebase. It uses Retrieval-Augmented Generation (RAG) to fetch canonical documentation and runnable code examples from the local knowledge source, and a Knowledge Graph (KG) to validate AI-generated code against the actual library structure. In practice, you can ask the navigator questions about LangGraph usage, and it will retrieve relevant docs, propose concrete code snippets, and verify them against the codebase before delivering a final, executable answer. The system emphasizes environment-aware, version-true responses, reducing hallucinations and improving first-pass success when coding against LangGraph.

Prerequisites

• Docker (recommended for local container-based setup) or a Python 3.12+ environment with uv (or uvx) tooling
• Python 3.12+ and uv: for local development and running the server
• API keys/credentials: OpenAI API Key for embeddings and LLM calls; a Supabase project for vector storage (if using the local knowledge server)

1. Clone the repository and submodules

• If you haven’t already cloned with submodules:

git clone --recursive https://github.com/botingw/langgraph-dev-navigator.git
cd langgraph-dev-navigator

• If you already cloned without submodules, fetch them:

git submodule update --init --recursive

2. Install core dependencies

• Install Python requirements (in a Python 3.12+ environment):

uv pip install -r requirements.txt

3. Configure environment variables

• Create a .env file (or export these in your shell):

OPENAI_API_KEY=your-openai-api-key
SUPABASE_URL=your-supabase-url
SUPABASE_ANON_KEY=your-supabase-anon-key

• You may also set any optional variables required by the local knowledge server or RAG components as documented in the repository.

4. Run the server

• If using the local UV/uvx path, start the server with:

uv run langgraph-dev-navigator

• Alternatively, if you’re using Docker, build and run the container as described in the Docker setup in the project docs.

5. Verify the server is running

• Open http://localhost:8000 (or the configured port) to ensure the MCP server is up, then begin issuing queries through your MCP client.

Tips and common issues:

• Ensure your API keys and credentials are correctly configured; missing keys are a common source of failures in RAG or embedding generation.
• If you encounter submodule issues after pulling updates, re-run the recursive submodule update commands.
• When validating code, rely on the Knowledge Graph checks to avoid deprecated or non-existent API usage.
• For hosted remote MCP clients, you can skip local backend setup and rely on the hosted URL shared by maintainers; use the Remote MCP Client Quickstart documentation referenced in the repo for that path.
• Environment variables can sometimes differ between local and hosted deployments; consult the README and environment docs for the exact variable names and recommended defaults.