QAService is a template for deploying a QA system using RAG (Retrieval Augmented Generation) technology.

To deploy the service you need to have Docker compose installed:

1. Upload documents to the documents folder
2. python create_chroma.py
3. Fill in the .env (at a minimum OPENAI_API_KEY)
4. Copy everything to the remote server
5. docker-compose up -d -build

Depending on how start.py is configured, you may be able to run

• UvicornService - RestAPI
• TG_Service - Telegram bot
• ST_Service - StreamLit

You can customize the RAG system by changing:

• The way of reading and splitting documents (by RecursiveCharacterTextSplitter)
• Choose a different VectorStore (default is ChormaDB)
• Choose a different Embedding model (default text-embedding-3-small)
• Select a different retriever method
• Select a different LLM (optional gpt-3.5-turbo)

Query information is logged to PosgresQL database, with PGAdmin interface.

Also based on TruLens and Ragas, tracking of system parameters is implemented:

• context_relevance, answer_relevance, groundedness
• context_precision, context_recall, faithfulness, answer_relevancy