Skip to content

CLAUDE.md

Latest commit

 

History

History
84 lines (61 loc) · 3.09 KB

CLAUDE.md

File metadata and controls

84 lines (61 loc) · 3.09 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

This is a Course Materials RAG (Retrieval-Augmented Generation) system - a full-stack web application that enables semantic search and AI-powered question answering over course documents. The system uses ChromaDB for vector storage, Anthropic's Claude for generation, and serves a web interface for user interaction.

Essential Commands

Running the Application

# Quick start (recommended)
./run.sh

# Manual start
cd backend && uv run uvicorn app:app --reload --port 8000

# Development with different port (if 8000 is occupied)
cd backend && uv run uvicorn app:app --reload --port 8001

Setup Commands

# Install dependencies
uv sync

# Create environment file (required)
cp .env.example .env
# Then edit .env with your ANTHROPIC_API_KEY

Architecture

Core Components

The system follows a modular RAG architecture with clear separation of concerns:

RAGSystem (backend/rag_system.py) - Main orchestrator that coordinates all components:

  • DocumentProcessor: Chunks course documents into searchable segments
  • VectorStore: Manages ChromaDB for semantic search using sentence transformers
  • AIGenerator: Handles Anthropic Claude API calls for response generation
  • SessionManager: Maintains conversation context and history
  • ToolManager: Coordinates search tools for enhanced retrieval

Data Models (backend/models.py):

  • Course: Contains title, instructor, lessons, and course metadata
  • Lesson: Individual lessons with titles and links
  • CourseChunk: Text segments with course/lesson context for vector storage

Frontend Architecture:

  • Vanilla HTML/CSS/JS with pink-themed UI (frontend/style.css)
  • JavaScript handles API communication and UI updates (frontend/script.js)
  • FastAPI serves both static files and API endpoints

Key Configuration (backend/config.py)

  • Uses Claude Sonnet 4 model (claude-sonnet-4-20250514)
  • Embedding model: all-MiniLM-L6-v2
  • Document chunks: 800 characters with 100 character overlap
  • Conversation history: 2 messages retained
  • ChromaDB path: ./chroma_db

API Endpoints

  • POST /api/query: Submit questions, returns AI response with sources
  • GET /api/courses: Retrieve course statistics and titles
  • /: Static file serving for frontend

Document Processing Flow

  1. Course documents (.txt files) placed in docs/ folder
  2. DocumentProcessor extracts course metadata and lessons from structured text
  3. Content is chunked and stored in ChromaDB with course/lesson context
  4. User queries trigger semantic search + Claude generation with retrieved context
  5. Session manager maintains conversation history for follow-up questions

Development Notes

  • Uses uv for Python dependency management
  • FastAPI with auto-reload for backend development
  • Frontend styling uses CSS custom properties for theming
  • ChromaDB persists to local filesystem for data retention
  • Requires ANTHROPIC_API_KEY environment variable