Pathfolio — AI-Powered Digital Identity System

Built for MemoryVerse AI '26

"I never have to search through folders again."

Pathfolio turns a student's scattered certificates, resumes, project reports, internship letters, and portfolio links into a single, intelligent digital identity — automatically organized, semantically connected, and instantly searchable in plain English.

---

What it does

Module	What it solves
1. AI Data Ingestion	Upload any certificate, resume, report, letter, or portfolio link. Original file/format is preserved via Cloudinary.
2. Intelligent Categorization	Every upload is auto-classified into Projects / Skills / Certifications / Internships / Achievements / Academics — no manual sorting.
3. Relationship Engine	A hybrid embeddings + LLM pipeline finds and labels connections: Certification → Skill → Project → Internship → Career Path.
4. Digital Journey Timeline	Every document is placed on a year-by-year visual timeline of your growth.
5. Smart Retrieval	Ask "show my AI projects" or "show my latest resume" in plain English and get the original files back instantly.

---

Tech Stack

• Frontend: React (Vite) + Tailwind CSS v4, React Router, Recharts/Lucide icons
• Backend: Python FastAPI
• LLM: Groq API (Llama 3.3 70B) — free tier, used for categorization, relationship labeling, and natural-language query parsing
• Embeddings: sentence-transformers (all-MiniLM-L6-v2) — runs locally, free, no rate limits
• Vector DB: ChromaDB (embedded, persisted locally)
• File storage: Cloudinary (free tier) — preserves original file format, gives instant shareable URLs
• Metadata DB: SQLite (via SQLAlchemy)
• OCR: Tesseract (pytesseract) for scanned certificates/images; pypdf + pdf2image for PDFs

Why this stack: everything except the Groq LLM call runs entirely in your own backend (embeddings, vector search, OCR) — so the only external dependency that needs a real API key is Groq, which is free.

---

Architecture

See docs/architecture.md for the full diagram and data-flow explanation.

Upload (file or link)
   │
   ▼
Text Extraction (pypdf / pytesseract OCR / python-docx)
   │
   ▼
Groq LLM ── classify category, extract title/summary/entities/year
   │
   ▼
Cloudinary ── stores original file, returns permanent URL
   │
   ▼
SQLite ── persists Document row (category, entities, year, url)
   │
   ▼
MiniLM Embedding ── vectorize (title+summary+category+entities)
   │
   ▼
ChromaDB ── stores vector for semantic search
   │
   ▼
Relationship Engine ── nearest-neighbour candidates (Chroma) → Groq confirms + labels edge
   │
   ▼
Timeline Service ── places document on the year timeline

Retrieval flow:

Natural language query → Groq parses intent (category/keywords/sort)
   → MiniLM embeds query → ChromaDB similarity search
   → SQLite fetch full records → return original file URLs

---

Project Structure

pathfolio/
├── backend/
│   ├── app/
│   │   ├── main.py                  FastAPI app entrypoint
│   │   ├── config.py                Settings from .env
│   │   ├── database.py              SQLAlchemy engine/session
│   │   ├── schemas.py               Pydantic response models
│   │   ├── models/models.py         Document, Relationship, TimelineEntry
│   │   ├── services/
│   │   │   ├── cloudinary_service.py
│   │   │   ├── extraction_service.py    PDF/OCR/DOCX text extraction
│   │   │   ├── groq_service.py          LLM categorization/relationships/query parsing
│   │   │   ├── embedding_service.py     MiniLM + ChromaDB
│   │   │   ├── relationship_engine.py   Hybrid embedding+LLM relationship builder
│   │   │   └── timeline_service.py
│   │   └── routers/
│   │       ├── upload.py            Module 1 + 2 (ingestion pipeline)
│   │       ├── documents.py         CRUD + category summary
│   │       ├── search.py            Module 5 (smart retrieval)
│   │       ├── timeline.py          Module 4
│   │       └── relationships.py     Module 3 (knowledge graph)
│   ├── requirements.txt
│   ├── .env.example
│   ├── Dockerfile
│   └── run.sh
├── frontend/
│   ├── src/
│   │   ├── api/client.js            Axios API client
│   │   ├── components/              Sidebar, DocumentCard
│   │   ├── pages/                   Dashboard, Upload, Timeline, Graph, Search
│   │   ├── App.jsx, main.jsx, index.css
│   ├── .env.example
│   └── package.json
└── docs/
    ├── architecture.md
    └── thought-process.md

---

Setup & Run Locally

1. Backend

cd backend
cp .env.example .env
# Edit .env and paste in your FREE API keys:
#   GROQ_API_KEY        -> https://console.groq.com/keys
#   CLOUDINARY_CLOUD_NAME / API_KEY / API_SECRET -> https://cloudinary.com/users/register/free

./run.sh
# or manually:
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --reload

Backend runs at http://localhost:8000 (interactive API docs at /docs).

System dependency: Tesseract OCR must be installed for scanned-image certificates to be readable:

# Ubuntu/Debian
sudo apt-get install tesseract-ocr poppler-utils
# Mac
brew install tesseract poppler

2. Frontend

cd frontend
cp .env.example .env   # points to http://localhost:8000 by default
npm install
npm run dev

Frontend runs at http://localhost:5173.

---

Deployment

• Backend: Deploy via the included Dockerfile to Render / Railway / Fly.io. Set the three env vars (GROQ_API_KEY, CLOUDINARY_*) in the platform's environment variable settings — never commit .env.
• Frontend: Deploy to Vercel/Netlify. Set VITE_API_BASE_URL to your deployed backend URL.
• ChromaDB and SQLite persist to disk — on ephemeral hosts (like Render's free tier) attach a persistent disk/volume, or swap SQLite for a managed Postgres and Chroma for a hosted vector DB if you need durability across restarts.

---

Free API Keys Needed

Service	Free tier	Where
Groq	Yes, generous free tier	https://console.groq.com/keys
Cloudinary	Yes, free tier (25 credits/mo)	https://cloudinary.com/users/register/free

Everything else (embeddings, vector DB, OCR) runs locally at no cost.

---

Demo Script (for judges)

1. Upload a certificate (PDF/image) — watch it get auto-classified into Certifications with extracted skills, in seconds.
2. Upload a project report and an internship letter.
3. Open Knowledge Graph — see the certificate's skill connected to the project, and the project connected to the internship.
4. Open Timeline — see all three placed on the correct year.
5. Go to Smart Search, type "show my AI projects" — the original file opens instantly, no folder browsing.