A Retrieval-Augmented Generation (RAG) system designed to answer questions about course materials using semantic search and AI-powered responses.

This application is a full-stack web application that enables users to query course materials and receive intelligent, context-aware responses. It uses ChromaDB for vector storage, Anthropic's Claude for AI generation, and provides a web interface for interaction.

The system implements a Retrieval-Augmented Generation (RAG) pipeline:

• Document Processing: Ingests course documents (TXT, PDF, DOCX) via backend/document_processor.py. Extracts metadata (title, instructor, links) and lessons using regex patterns. Chunks content into ~800-character segments with 100-character overlap, enriching with course/lesson context.
• Vector Storage: ChromaDB (./chroma_db) stores embeddings using Sentence Transformers ("all-MiniLM-L6-v2"). Separate collections for course metadata (course_catalog) and content chunks (course_content).
• Query Processing: FastAPI backend (backend/app.py) handles /api/query. Orchestrated by RAGSystem (backend/rag_system.py), which uses tools for semantic search and Claude for synthesis.
• AI Integration: Anthropic Claude (model: "claude-sonnet-4-20250514") via AIGenerator (backend/ai_generator.py). Supports tool use (e.g., CourseSearchTool for retrieval) and session history (up to 2 exchanges).
• Frontend: Static HTML/JS/CSS (frontend/) with chat UI, markdown rendering, and API calls.
• Key Features: Semantic search with filters (course/lesson), sourced responses, conversational sessions.

Documents must follow a structured format for optimal processing:

• Line 1: Course Title: [Title]
• Line 2: Course Link: [URL] (optional)
• Line 3: Course Instructor: [Name] (optional)
• Subsequent: Lesson X: [Title] followed by content; optional Lesson Link: [URL].

Sample docs in docs/ cover AI/LLM topics (e.g., Anthropic courses).

• Python 3.13 or higher
• uv (Python package manager)
• An Anthropic API key (for Claude AI)
• For Windows: Use Git Bash to run the application commands - Download Git for Windows

1. Install uv (if not already installed)

   curl -LsSf https://astral.sh/uv/install.sh | sh

2. Install Python dependencies

   uv sync

3. Set up environment variables

   Create a .env file in the root directory:

   ANTHROPIC_API_KEY=your_anthropic_api_key_here

Use the provided shell script:

chmod +x run.sh
./run.sh

cd backend
uv run uvicorn app:app --reload --port 8000

The application will be available at:

• Web Interface: http://localhost:8000
• API Documentation: http://localhost:8000/docs

• Root: Configuration (pyproject.toml, .env.example, run.sh).
• backend/: FastAPI app and core logic (8 Python modules: app.py, config.py, rag_system.py, etc.).
• frontend/: Static web UI (index.html, script.js, style.css).
• docs/: Sample course transcripts (4 TXT files).
• ./chroma_db: Auto-created vector database (delete to reload docs).

• Sessions: Limited to 2 history exchanges for context; managed via session_manager.py.
• Search Limits: Top-5 results per query; one search max via tool use.
• Customization: Add docs to docs/ (restart to reload). Extend tools in search_tools.py.
• Dependencies: Locked via uv.lock; uses uv for virtual env management.

• API Key Issues: "Invalid API key" or 500 errors → Verify .env and restart. Ensure key has Claude access.
• No Courses Loaded: Empty stats/UI → Check docs/ files; delete ./chroma_db and restart for reload. Verify processing logs.
• Port Conflicts: "Address in use" → Change --port 8001 in run command.
• Windows: Use Git Bash; ensure uv in PATH (echo $PATH).
• Query Errors: "No content found" → Docs may not match query; test with samples (e.g., "MCP course outline").
• High Latency: Long queries → Claude token limits; check API usage in Anthropic console.
• Clean Start: rm -rf ./chroma_db then restart to rebuild vectors.

For production: Consider Docker, environment-specific configs, or cloud vector DB (e.g., Pinecone).