Cloud Code Agentic RAG Masterclass

Build an agentic RAG application from scratch by collaborating with Claude Code. Follow along with our video series using the docs in this repo.

Watch the full video on YouTube

What This Is

A hands-on course where you collaborate with Claude Code to build a full-featured RAG system. You're not the one writing code—Claude is. Your job is to guide it, understand what you're building, and course-correct when needed.

You don't need to know how to code. You do need to be technically minded and willing to learn about APIs, databases, and system architecture.

What You'll Build

• Chat interface with threaded conversations, streaming, tool calls, and subagent reasoning
• Document ingestion with drag-and-drop upload, multi-format parsing, delete controls, and processing status
• Full RAG pipeline: chunking, embedding, hybrid search, reranking
• Agentic patterns: text-to-SQL, web search, subagents with isolated context

Tech Stack

Layer	Tech
Frontend	React, TypeScript, Tailwind, shadcn/ui, Vite
Backend	Python, FastAPI
Database	Supabase (Postgres + pgvector + Auth + Storage)
Doc Processing	Docling + native text fallbacks
AI Models	Local (LM Studio) or Cloud (OpenAI, OpenRouter)
Observability	LangSmith

The 8 Modules

1. App Shell — Auth, chat UI, managed RAG with OpenAI Responses API
2. BYO Retrieval + Memory — Ingestion, pgvector, switch to generic completions API
3. Record Manager — Content hashing, deduplication
4. Metadata Extraction — LLM-extracted metadata, filtered retrieval
5. Multi-Format Support — PDF, DOCX, HTML, Markdown via Docling
6. Hybrid Search & Reranking — Keyword + vector search, RRF, reranking
7. Additional Tools — Text-to-SQL, web search fallback
8. Subagents — Isolated context, document analysis delegation

Getting Started

1. Clone this repo
2. Install Claude Code
3. Open in your IDE (Cursor, VS Code, etc.)
4. Run claude in the terminal
5. Use the /onboard command to get started

Docs

• PRD.md — What to build (the 8 modules in detail)
• CLAUDE.md — Context for Claude Code
• PROGRESS.md — Track your build progress

Local Module 1 Scaffold

The repository now includes a local Module 1 implementation scaffold:

• backend/ FastAPI API with JWT auth, multi-format document ingestion, document deletion, pgvector retrieval, SSE chat streaming, realtime ingestion status streams, and optional LangSmith tracing
• frontend/ Vite + React client with auth, separate Chat/Ingestion interfaces, drag-and-drop upload, live document status updates, conversation list, and grounded chat UI
• docker-compose.yml for local PostgreSQL + pgvector
• .env.example plus frontend/.env.example for environment bootstrap
• Module 7 chat orchestration that can combine document retrieval, workspace text-to-SQL over a user-scoped SQLite snapshot, and optional web-search fallback with source attribution
• OCR fallback for scanned PDFs through Docling when the initial PDF extraction returns too little text
• Database bootstrap now provisions a non-bypass application role and the app connections SET ROLE into it so the Postgres RLS policies are actually enforced during runtime

Run It

1. Copy .env.example to .env and set your local LLM endpoint values. Module 7 notes:
   • SQL_TOOL_ROW_LIMIT controls how many structured rows are passed into the final answer prompt.
   • WEB_SEARCH_ENABLED defaults to true; set it to false if you want to disable outbound web fallback.
   • WEB_SEARCH_PROVIDER=duckduckgo_html works without an API key.
   • WEB_SEARCH_PROVIDER=tavily requires WEB_SEARCH_API_KEY. OCR notes:
   • PDF_OCR_ENABLED=true enables automatic OCR fallback for low-text PDFs.
   • PDF_OCR_ENGINE=tesseract_cli expects the tesseract CLI to be installed and available on your PATH.
   • For French scanned PDFs, PDF_OCR_LANGUAGES=fra,eng is a reasonable starting point.
   • If you switch to easyocr, use engine-appropriate language codes such as fr,en.
2. Start PostgreSQL with docker compose up -d.
3. Create a backend venv, install backend/requirements.txt for Docling-backed parsing support, then run python backend/scripts/init_db.py. This applies the schema migrations and provisions the non-bypass rag_app database role used for RLS-enforced app traffic.
4. Start the backend from backend/ with uvicorn app.main:app --reload.
5. Install frontend dependencies in frontend/ and run npm run dev.

Test Account

Use this local account for manual smoke testing once the app is running:

• Email: test@test.com
• Password: Test123456!

Current confirmed local test password: Test123456!

Python note: the current OpenAI SDK emits an upstream compatibility warning on Python 3.14, so Python 3.12 or 3.13 is the safer local target if you want a quieter runtime.

Join the Community

If you want to connect with hundreds of builders creating production-grade AI and RAG systems, join us in The AI Automators community. Share your progress, get help when you're stuck, and see what others are building.