Skip to content

CLAUDE.md

Latest commit

 

History

History
84 lines (61 loc) · 2.78 KB

CLAUDE.md

File metadata and controls

84 lines (61 loc) · 2.78 KB

CLAUDE.md

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

Commands

Run the application:

./run.sh
# Or manually:
cd backend && uv run uvicorn app:app --reload --port 8000

Install dependencies:

uv sync

Access points when running:

Architecture

This is a RAG (Retrieval-Augmented Generation) chatbot for course materials. The system uses ChromaDB for vector storage and Claude API for response generation.

Request Flow

User Query → FastAPI (app.py) → RAGSystem.query()
                                     ↓
                              AIGenerator (Claude API with tools)
                                     ↓
                              CourseSearchTool → VectorStore.search()
                                     ↓
                              ChromaDB semantic search
                                     ↓
                              Results → Claude synthesizes response

Key Components (backend/)

  • RAGSystem (rag_system.py): Main orchestrator that wires together all components. Entry point for queries via query() and document ingestion via add_course_folder().

  • VectorStore (vector_store.py): ChromaDB wrapper with two collections:

    • course_catalog: Course metadata for name resolution
    • course_content: Chunked content for semantic search
    • Key method: search(query, course_name, lesson_number) handles course name resolution then content search.

  • AIGenerator (ai_generator.py): Claude API integration with tool-calling support. Handles the agentic loop: initial request → tool execution → follow-up response.

  • DocumentProcessor (document_processor.py): Parses course documents with expected format (Course Title/Link/Instructor in first lines, then "Lesson N:" markers). Uses sentence-aware chunking with configurable overlap.

  • ToolManager/CourseSearchTool (search_tools.py): Tool definitions for Claude's tool-use. CourseSearchTool wraps VectorStore and tracks sources.

Document Format

Course documents in docs/ must follow:

Course Title: [title]
Course Link: [url]
Course Instructor: [name]

Lesson 0: [lesson title]
Lesson Link: [url]
[content...]

Lesson 1: [lesson title]
[content...]

Configuration

All settings in backend/config.py, loaded from .env:

  • ANTHROPIC_API_KEY: Required
  • ANTHROPIC_MODEL: Default claude-sonnet-4-20250514
  • CHUNK_SIZE: 800 chars
  • CHUNK_OVERLAP: 100 chars
  • MAX_RESULTS: 5 search results
  • CHROMA_PATH: ./chroma_db

Data Persistence

ChromaDB data stored in backend/chroma_db/. On startup, app.py loads documents from docs/ folder, skipping already-indexed courses.