smart chat

This is Till's version of the Chat Langchain.

The original Chat Langchain is a chatbot specifically focused on question answering over the LangChain documentation. Built with LangChain, FastAPI, and Next.js.

From there I modified it (and plan to modify it) in the following ways:

• Running locally with Ollama ✅
• Offering multiple local models
• Offering different sources: Another website (maybe a manual of another software), a Wiki space...
• Providing functionality to evaluate performance, i.e. if the answers are close to what we expect.

Todo / backlog

• Find out ifthe _scripts directory is needed.
• Server component should list the available models so client offers option to choose
• Server component offers different brainz, i.e. sites that have been scraped and ingested

Done

• 2024-04-25 Renamed Github Repo
• 2024-04-24 Move from poetry to requirements file - as poetry is too complex for a hobby programmer like me.
• 2024-04-22 Simplify the original project and get it running fully locally

Running locally

1. Create a local Python environment: python3.12 -m venv .venv
2. Activate it with source .venv/bin/activate
3. Install backend dependencies: pip install -r requirements.txt.
4. Makle sure Ollama is running and has the provided models pulled. You can check it with ollama list.
5. Configure your constants in backend/constants.py
6. Run python backend/ingest.py to ingest LangChain docs data into the Chroma vectorstore (only needs to be done once).
7. Start the Python backend with python backend/main.py.
8. Install frontend dependencies by running cd ./frontend, then yarn.
9. Run the frontend with yarn dev for frontend.
10. Open localhost:3000 in your browser.