A unified AI platform for Stony Brook University, built for every stage of the Seawolf journey.

localhost:3000   Seaport Portal       — unified landing page for all apps
localhost:3001   Admin Dashboard      — content management & quality monitoring
localhost:3002   SB-lumni             — alumni matching & community
localhost:3003   StudyCoach           — AI-powered Socratic tutor
localhost:8000   FastAPI Backend      — shared API for all apps

RAG-powered Q&A chatbot grounded in official SBU data.

• 22,000+ chunks crawled from stonybrook.edu, with source citations on every answer
• Intent classification → pgvector retrieval → authority reranking → LLM synthesis
• Office routing when a question needs a real person
• Confidence scoring, topic browsing, conversation history

Full content management and quality monitoring interface for platform operators.

• Source management — add, edit, or toggle the URLs the AI learns from; set authority scores and categories per source
• Crawl & reindex jobs — trigger fresh data fetches and re-embeddings directly from the UI, no code needed
• Chunk inspector — browse every ingested page and drill into the individual vector chunks
• FAQ overrides — curate high-priority Q&A pairs that bypass RAG entirely for guaranteed-accurate answers
• Conversation review — read every chat session with full message history, confidence scores, and citations
• Evaluation runner — run structured SBU Q&A test suites and see pass/fail rates and scores per case

AI-powered alumni matching and community for Stony Brook graduates.

• Restricted to @stonybrook.edu / @alumni.stonybrook.edu — enforced via JWT domain check
• 3-step onboarding with optional résumé upload (LLM-parsed)
• Two-stage matching pipeline — ANN retrieval via pgvector → multi-signal reranking (major, career path, skills Jaccard, graduation proximity, MMR diversity λ=0.7)
• Connect system, community feed with hashtags, comments, and likes

Socratic AI tutor built around a student's own course materials.

• Upload PDFs, DOCX, PPTX, or code files — AI auto-generates a learning map with meaningful section titles
• Parsing pipeline: page extraction → concept grouping (~600 words) → LLM title generation
• Teach mode — Socratic system prompt: one concept per message, one question at the end, never gives the answer directly
• Session history, resume or restart any past study session, study plan with progress tracking

┌───────────────────────────────────────────────────────────────┐
│  Seaport Portal :3000  │  Admin :3001  │  SB-lumni :3002      │
│  StudyCoach :3003          (Next.js App Router + TailwindCSS) │
├───────────────────────────────────────────────────────────────┤
│  FastAPI Backend :8000                                        │
│  ┌──────────┬───────────┬────────────┬──────────────────────┐ │
│  │ /public  │  /admin   │  /alumni   │  /studycoach         │ │
│  │ RAG Q&A  │  Mgmt+Eval│  Matching  │  Courses + Teaching  │ │
│  └──────────┴───────────┴────────────┴──────────────────────┘ │
├───────────────────────────────────────────────────────────────┤
│  PostgreSQL + pgvector  │  Redis  │  File Storage (local/S3) │
└───────────────────────────────────────────────────────────────┘

JWT auth restricted to @stonybrook.edu and @alumni.stonybrook.edu domains.

• Docker Desktop
• Git

git clone https://github.com/gdakate/ask-seawolf.git
cd ask-seawolf

cp .env.example .env

docker compose up -d --build

All four apps and the API start together. Optionally seed demo data:

docker compose exec api python seed_alumni.py   # 50+ alumni profiles
docker compose exec api python seed_posts.py    # feed posts + connections

For Ask Seawolf and StudyCoach, register with any @stonybrook.edu address.

Set AI_PROVIDER in .env:

ask-seawolf/
├── apps/
│   ├── web/            # Seaport portal + Ask Seawolf chat (:3000)
│   ├── admin/          # Admin dashboard (:3001)
│   ├── alumni/         # SB-lumni frontend (:3002)
│   ├── studycoach/     # StudyCoach frontend (:3003)
│   └── api/            # FastAPI backend (shared, :8000)
│       ├── app/
│       │   ├── routers/
│       │   │   ├── public.py       # Ask Seawolf RAG endpoints
│       │   │   ├── admin.py        # Admin CRUD, crawl, eval
│       │   │   ├── alumni.py       # SB-lumni matching + feed
│       │   │   └── studycoach.py   # Courses, materials, sessions, teach
│       │   ├── models/             # SQLAlchemy ORM models
│       │   ├── services/           # RAG pipeline, matching, AI providers
│       │   └── core/               # Config, DB session, JWT auth
│       ├── migrations/             # Alembic (001–006)
│       ├── seed_alumni.py
│       └── seed_posts.py
├── infra/terraform/    # AWS ECS/Fargate infrastructure
├── docs/               # Architecture, deployment, RAG pipeline, video script
├── docker-compose.yml
└── .env.example

docker compose up -d --build    # Start everything
docker compose logs -f          # Follow logs
docker compose exec api pytest  # Run backend tests
open http://localhost:8000/docs # API docs (Swagger)
docker compose exec api alembic upgrade head  # Run migrations

Full list in .env.example.

AWS infrastructure (ECS Fargate + RDS + ElastiCache) is defined in infra/terraform/. See docs/deployment.md for step-by-step instructions.