readmenet

This repo implements a chatbot that can interact with users and provide answers to their queries about their own data. It uses language models, text splitters, embeddings, vector stores, and conversational retrieval chains to enable the chatbot's functionality. The code also includes features for loading and managing documents, as well as logging chat history. Users can input queries, clear the chat history, load documents into the vector store, and delete stored documents. The code aims to create a conversational interface for retrieving custom information and interacting with the chatbot.

Getting started

1. Download and install Docker Desktop

2. Git checkout this repo, then cd /path/to/repo/readmenet

3. Generate an openAI API key and put it into a new secrets.env file in readmenet/ containing:

OPENAI_API_KEY=someapikeyhere

4. mkdir data && mkdir data/docs

5. docker compose up (note: use docker compose build to force rebuilds if you edit your dockerfile, etc)

6. Browse to http://localhost:7860/

7. Place documents into readmenet/data/docs - they will show up under /app/data/docs in the tools 'Load docs' tab

8. Note: you can save/reuse/change-out chromadb databases without recalculating the vector embeddings by making cp -r copies of the relevant autogenerated chromadb directories in readmenet/data/chroma

Caveat emptor: This is alpha/pilot code, with breaking changes expected along the way. Improvements, fixes, and enhancements by pull requests & issues welcome.