RAG Advisor
NavyDevilDoc/rag-advisor
Wizard de 12 perguntas para escolher a melhor arquitetura RAG
Ver no GitHubUm agente interativo em forma de wizard com 12 perguntas que recomenda a arquitetura RAG ideal (Padrão Vetorial, Grafo ou Agêntica) para seu caso de uso específico, com raciocínio avaliado por IA e guia de paradigma de deploy. Combina interface React com backend FastAPI e API Anthropic.
Instalação
npx convez-ai NavyDevilDoc/rag-advisor1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Decidir entre arquiteturas RAG (vetorial, grafo, agêntica)
- Validar escolhas de design antes de implementar pipelines
- Treinar times em padrões de RAG com recomendações IA-assistidas
Veredito

Ferramenta sólida de decisão para quem está no ponto de inflexão arquitetural em projetos RAG. Bem-estruturada tecnicamente (Vite+React+FastAPI), pronta para produção no Railway, com raciocínio IA explícito. Valioso para reduzir riscos em escolhas de stack complexas.
README
A tradução para o português deste README ainda está na fila.
RAG Architecture Advisor
A 12-question wizard that recommends the optimal RAG architecture (Standard Vector / Graph / Agentic) for a given use case, with deployment paradigm guidance and AI-generated reasoning from the Anthropic API.
Stack
- Frontend: Vite + React 18, inline styles, lucide-react icons
- Backend: FastAPI + Uvicorn, Anthropic Python SDK
- Deploy: Railway single service (FastAPI serves built Vite static files)
See CLAUDE.md for full architecture details and design rationale. See PROGRESS.md for build status.
Prerequisites
- Node.js 18+
- Python 3.11+
- uv — fast Python package/project manager
- An Anthropic API key
First-time setup
# Frontend
cd frontend
npm install
cp .env.example .env # optional in dev; Vite proxy handles /api/*
# Backend
cd ../backend
uv venv # creates .venv/
uv pip install -r requirements.txt # installs into .venv/
cp .env.example .env
# then edit backend/.env and set ANTHROPIC_API_KEY=sk-ant-...
Running locally
Two terminals:
# Terminal 1 — backend on :8000
cd backend
.venv/Scripts/uvicorn main:app --reload --port 8000 # Windows
# .venv/bin/uvicorn main:app --reload --port 8000 # macOS/Linux
# Terminal 2 — frontend on :5173
cd frontend
npm run dev
Or one terminal (from project root):
npm install # installs concurrently once
npm run dev # runs both
npm run dev invokes the venv's uvicorn directly so you don't need to
activate the venv manually. (The dev script is Windows-pathed; if you move
to macOS/Linux, switch .venv/Scripts/ → .venv/bin/ in package.json,
or add a minimal pyproject.toml so uv run uvicorn ... works on every OS.)
Vite dev server proxies /api/* to http://localhost:8000, so the frontend code
can call fetch('/api/analyze', ...) without worrying about CORS.
Project structure
See CLAUDE.md § Project Structure for the canonical tree and per-file responsibilities.
Deploy to Railway
- Push to GitHub and connect the repo to Railway.
- In the Railway service settings, add the environment variable:
ANTHROPIC_API_KEY=sk-ant-... - Railway uses railway.json — single service, FastAPI serves
built frontend at
/and API at/api/*. Healthcheck on/health.