plain-rag/README.md

PlainRAG

PlainRAG is RAG application without LLM orchestration frameworks like Langchain or Haystack but with LiteLLM, Transformers, FastAPI.

Quick Start

0. You need to install uv first
1. Copy .env.example to .env and fill in your values.
2. Run the following command to build and start the services:

docker compose up db -d
make install-deps # or uv sync
make create-tables
make start

Then navigate to http://127.0.0.1:8000/docs

or

docker compoes up -d

This will use Docker Compose to start the API and database services.

Environment Variables

• DB_SERVER: Database server hostname (default: localhost)
• DB_USER: Database username (default: postgres)
• DB_PASSWORD: Database password
• DB_NAME: Database name (default: chat_hub)
• DB_PORT: Database port (default: 5432)

Other variables are documented in .env.example.

Components

• API (app/api/endpoints.py): FastAPI endpoints for file ingestion, querying, and configuration.
• Services (app/services/):
  • rag_service.py: Core RAG pipeline (ingest, query, stream answers).
  • config_service.py: Manages model and vector store selection.
  • embedding_providers.py: Embedding model integration (e.g., MiniLM).
  • rerankers.py: Reranker model integration (e.g., CrossEncoder).
  • vector_stores.py: Vector database integration (PostgreSQL/pgvector).
• Core (app/core/):
  • config.py: Application settings and environment management.
  • registry.py: Registry pattern for models and stores.
  • utils.py: Text splitting utilities.
  • exception.py: Custom exceptions.
  • interfaces.py: Abstract interfaces for models and stores.
• Schemas (app/schemas/):
  • models.py: Pydantic models for API requests/responses.
  • enums.py: Enum definitions for model/store selection.