Backend Architecture (FastAPI)

Overview

The backend is a FastAPI application located in backend/. It powers the chatbot and RAG functionality.

Entry Point

• File: backend/main.py
• Run: uvicorn backend.main:app --reload (or via npm run dev)
• Port: Defaults to 8000.

Endpoints

POST /api/chat

• Purpose: Main RAG chat endpoint.
• Input: ChatRequest (query, history, user_context).
• Process:
  1. Embed query.
  2. Search Qdrant (search_qdrant).
  3. Build prompt (build_rag_prompt).
  4. Generate Agent response.
• Output: ChatResponse (answer, contexts).

POST /api/ask-selection

• Purpose: Targeted Q&A on selected text.
• Input: AskSelectionRequest (question, selected_text).
• Process:
  1. Validates selection length.
  2. Builds selection-specific prompt.
  3. specific Agent instructions.

Dependencies & Utils

• backend/utils/config.py: Qdrant initialization.
• backend/utils/helpers.py: Embedding and Prompt building logic.
• backend/models.py: OpenAI/Gemini client setup.

Environment Variables

• GEMINI_API_KEY: For LLM and Embeddings.
• QDRANT_URL, QDRANT_API_KEY: Vector DB connection.