llm-embed-qa

Question answering system built with vector dbs and LLMs.

Part of the content is implemented with reference to michaelliao/llm-embedding-sample

Requirements (Default Component)

• Python 3.10
• Docker

Install

1. Clone this repo

2. Install requirements with pip install -r requirements.txt

3. Startup PostgreSQL with Docker

   docker run -d \
      --rm \
      --name pgvector \
      -p 5432:5432 \
      -e POSTGRES_PASSWORD=password \
      -e POSTGRES_USER=postgres \
      -e POSTGRES_DB=postgres \
      -e PGDATA=/var/lib/postgresql/data/pgdata \
      -v /path/to/llm-embedding-qa/pg-data:/var/lib/postgresql/data \
      -v /path/to/llm-embedding-qa/pg-init-script:/docker-entrypoint-initdb.d \
      ankane/pgvector:latest

   NOTE: replace /path/to/... with real path.

4. Run python main.py, edit config.yaml to set your api_key of OpenAI.

5. Put your markdown format documents in docs folder.

   • There are the wiki files of QChatGPT in docs_examples folder.

6. Run python main.py again, it will automatically build the vector database and start the server.

Usage

• GET /ask
  • content: the content of the question
  • strict: (Optional) skip LLM request if strict=true and no related answer found in vector db