← Voltar a studyAI — Documentação do Projeto
📋 Visão Geral — studyAI
studyAI — Documentação do ProjetoApresentação
studyAI
AI-powered study platform — Generate structured learning material, practice questions, and get feedback from your own documents. Any domain, any topic.
Overview
studyAI turns your documents into a personalized study system. Upload PDFs and notes, define what you want to learn, and the system generates structured content, practice questions, and answers your doubts—all grounded in your material.
A carregar diagrama…
Value Proposition
| User | Use case |
|---|---|
| Professional | Upload guidelines, articles. "I want to study X." → Structured material + practice questions + Q&A. |
| Student | Upload notes, legislation. "Develop on civil liability." → Organized content with evaluation. |
| Interview prep | Upload job description + docs. "Create material on RAG." → Study, practice, clarify via chat. |
Differentiator: Content is grounded in your documents. No generic curricula—your knowledge base, your study path.
User Journey
A carregar diagrama…
Architecture
A carregar diagrama…
| Layer | Technology |
|---|---|
| Frontend | Next.js (App Router) |
| Backend | FastAPI |
| Auth | JWT + refresh tokens (own backend) |
| DB | PostgreSQL |
| Vector | pgvector |
| Storage | S3-compatible (MinIO, R2) |
| Queue | Celery + Redis |
| LLM | OpenAI |
| Orchestration | LangGraph |
Agent Graph
A carregar diagrama…
| Agent | Role |
|---|---|
| Router | Classifies intent, routes to the right agent |
| Content Planner | Plans structure from RAG retrieval |
| Content Writer | Writes sections grounded in chunks |
| Question Generator | Creates practice questions |
| Question Evaluator | Scores answers, gives feedback |
| RAG QA Agent | Answers questions from your material |
RAG Pipeline
3 estágios: Hybrid search (dense + BM25) → RRF (fusão por rank) → Cross-encoder rerank.
A carregar diagrama…
- Scope: One index per project. Isolation by
project_id. - Chunking: 500–1200 chars, paragraph/header boundaries.
- Retrieval: Dense + BM25 → RRF → cross-encoder → top-5 a 20 per agent.
Implementation Phases
| Phase | Focus | Duration |
|---|---|---|
| 0 — Foundation | Auth, DB, Docker | 2–3 wk |
| 1 — Core | Projects, upload, ingest, RAG | 2–3 wk |
| 2 — Agents | Router, Planner, Writer, QA, Evaluator | 3–4 wk |
| 3 — Learning | SM-2, user profile, memory | 2–3 wk |
| 4 — Production | Rate limit, cost control, deploy | 1–2 wk |
Documentation
| Document | Description |
|---|---|
| ARCHITECTURE | Layers, components, data flow |
| API_SPEC | REST API specification |
| DATA_MODEL | Database schema |
| AGENTS | Agent graph, intents, RAG per agent |
| RAG_DESIGN | Chunking, indexing, retrieval |
| FLOWS | User flows, sequences |
| ROADMAP | Sprints, timeline, checklists |
| DIAGRAMS | Index of all diagrams |
Full index: docs/INDEX.md
Quick Start (planned)
# Clone
git clone https://github.com/your-org/studyAI.git
cd studyAI
# Dev
docker-compose up -d
# API: http://localhost:8000
# Web: http://localhost:3000
References
Zona de prática
Sem perguntas. Clica em Editar para adicionar.