Code Labyrinth

Code Labyrinth is a tool for using LLM chatbots in coding workflows. It uses Langchain and PGVector to provide context about the working codebase and relevant libraries.

Features

1. Ingest document collections from any readthedocs website
2. Ingest all code files from a local directory or github repo and keep them up to date
3. Add relevant documents to the context using vectorstore similarity retrieval of questions asked to a LLM.
4. Choose the appropriate agent for the task - simple Question-Answer agents or Plan-and-Execute agent
5. Operate everything with a convenient user interface

Setup instructions

1. Start the Postgres PGVector server: docker compose up
2. Set up PGVector extension: python -c "from coder.db import install_extension; install_extension()"
3. Set up python environment: virtualenv venv && venv/bin/activate && pip install .
4. Start backend server: python -m coder api
5. Set up frontend environment: cd ui && npm install
6. Start frontend (dev) server: npm run start