SmartDocs AI

Enterprise RAG SaaS platform for secure document intelligence, with a public product page, one-click guest demo, document upload, retrieval, citations, tenant isolation, credits, usage logs, and technical review flow.

业务价值

Shows an AI-native SaaS product beyond a chatbot: workspace RBAC, document processing, RAG answers with sources, billing discipline, and recruiter-friendly guest demo access.

核心功能

Public product landing pageGuest demo loginRead-only reviewer modeMulti-tenant workspace RBACPDF/DOCX/TXT/MD uploadCelery document indexingHybrid retrieval with RRF

Next.js App RouterTypeScriptFastAPISQLAlchemyPostgreSQL/pgvectorRedis/CeleryLangGraphDeepSeek/Qwen-ready

线上演示 GitHub

overview

SmartDocs AI is the flagship enterprise RAG SaaS project, built to demonstrate a complete document intelligence workflow from landing page to guest demo, citations, credits, usage logs, and technical review.

problem

Enterprise AI demos often stop at a chatbot and do not show tenant isolation, document processing, observability, billing, or source-grounded answers.

solution

Build a production-style SaaS flow with a real public entry point, one-click guest routing, workspace auth, document upload, indexing, retrieval, streamed answers, source citations, debug visibility, and credit-safe usage logging.

key features

Public product landing page / Guest demo login / Read-only reviewer mode / Multi-tenant workspace RBAC / PDF/DOCX/TXT/MD upload / Celery document indexing / Hybrid retrieval with RRF / Streaming RAG answers / Source citations / Retrieval Debug Panel / Atomic credit billing / Usage logs / Expanded technical review page

tech stack

Next.js App Router / TypeScript / FastAPI / SQLAlchemy / PostgreSQL/pgvector / Redis/Celery / LangGraph / DeepSeek/Qwen-ready / Docker / Vercel

architecture

Next.js handles the app UI, FastAPI exposes layered routers and services, PostgreSQL/pgvector stores workspaces and chunks, Redis/Celery processes documents, and the RAG service streams answer events.

AI-assisted workflow

The demo path is clearly labeled demo-local until real DeepSeek or Qwen keys are configured, so reviewers can test the product loop without fake provider claims.

challenges and what I learned

The key learning was connecting RAG architecture to product constraints: RBAC, workspace isolation, source citations, atomic credit deduction, failure-safe logging, and deployment reality.

screenshots

Screenshots are planned for README and portfolio polish.