DocSmith

AI-powered recruitment platform for resume screening, candidate matching, and hiring assistance.

Extracts structured data from resumes using OCR + LLMs, stores them with vector embeddings for semantic search, and matches candidates against job descriptions.

---

Features

Category	Features
Document Processing	PDF/image upload, Llama Scout OCR, batch processing, field extraction, verification workflow
Search & Matching	Semantic search (Gemini embeddings), hybrid search (BM25 + vector), JD matching, natural language queries
AI Assistants	Shortlist refinement (filter/reset/undo), hiring analysis & recommendations, profile chat
Platform	Analytics dashboard, JWT auth, role-based access (recruiter/job_seeker), profile management

---

Tech Stack

Layer	Technologies
Backend	FastAPI, Python 3.10+, MongoDB Atlas, LangChain, LangGraph
AI/ML	Groq (Llama Scout 17B OCR, Llama 3.3 70B LLM), Gemini (embeddings)
Frontend	React 18, TypeScript, Vite 5, Tailwind CSS, shadcn/ui, TanStack Query
Auth	PyJWT + bcrypt, role-based access
Search	BM25 + cosine similarity hybrid ranking

---

Architecture

  Frontend (React + Vite)  <--- REST --->  Backend (FastAPI)

  - Dashboard                               Multi-Agent Pipeline
  - Search                                  Document -> OCR -> Extract ->
  - Upload                                  Embed -> Store
  - JD Matching
  - Chat                                    AI: Groq (Llama) + Gemini
                                            DB: MongoDB (Documents + Vectors)

Processing Pipeline:

Upload -> Classification -> OCR Extraction -> Field Extraction -> Embedding -> MongoDB

---

Prerequisites

Requirement	Version
Python	3.10+
Node.js	18+
MongoDB	6.0+ (Atlas recommended)
Poppler	Latest

API Keys

Service	Purpose	Link
Groq	OCR + LLM	console.groq.com
Google Gemini	Embeddings	aistudio.google.com

---

Quick Start

1. Clone

git clone <repository-url>
cd job-profile-screening

2. Backend

cd backend
python -m venv venv
venv\Scripts\activate          # Windows
source venv/bin/activate       # Linux/Mac
pip install -r requirements.txt
cp .env.example .env           # Edit with your API keys
uvicorn src.main:app --reload --port 8000

3. Frontend

cd frontend
npm install
npm run dev

4. Verify

Service	URL
Frontend	http://localhost:8080
Backend API	http://localhost:8000
API Docs	http://localhost:8000/docs
Health Check	http://localhost:8000/health

---

Docker

# Build and run
docker-compose up --build

# Stop
docker-compose down

Deploy to Render

1. Push code to GitHub
2. Create Web Service on Render → connect repo → Root Directory: backend → Runtime: Docker
3. Set environment variables in Render dashboard:

   GROQ_API_KEY=your_key
   GEMINI_API_KEY=your_key
   MONGO_DB_URI=mongodb+srv://...
   JWT_SECRET=your_secret
   

---

Environment Variables

Backend (backend/.env)

Variable	Required	Description
GROQ_API_KEY	Yes	Groq API key for Llama models
GEMINI_API_KEY	Yes	Google Gemini API key
MONGO_DB_URI	Yes	MongoDB connection string
JWT_SECRET	Yes	JWT signing secret (min 32 chars)
LOG_LEVEL	No	DEBUG, INFO, WARNING, ERROR (default: INFO)

Frontend (frontend/.env.local)

Variable	Default	Description
VITE_API_BASE	http://localhost:8000	Backend API URL

---

API Overview

All endpoints prefixed with /api/v1. Full docs at /docs.

Auth

Method	Endpoint	Description
POST	/auth/signup	Register user
POST	/auth/login	Login, get JWT

Resumes

Method	Endpoint	Description
POST	/resumes/upload	Upload resume
POST	/resumes/upload-batch	Batch upload
GET	/resumes/	List profiles (paginated)
GET	/resumes/{id}	Get profile
GET	/resumes/search/hybrid	Hybrid search
PATCH	/resumes/{id}/fields	Update fields (draft)
POST	/resumes/{id}/verify	Verify and index
DELETE	/resumes/{id}	Delete profile

Job Descriptions

Method	Endpoint	Description
POST	/jd/upload	Upload JD, get matches
GET	/jd/	List JDs
GET	/jd/{id}/matches	Get matches
DELETE	/jd/{id}	Delete JD

Dashboard

Method	Endpoint	Description
GET	/dashboard/summary	Full analytics
GET	/dashboard/skills	Skills breakdown
GET	/dashboard/experience	Experience distribution
GET	/dashboard/domains	Domain stats

Chat

Method	Endpoint	Description
POST	/chat/shortlist-assist	Refine shortlist
POST	/chat/hiring-assist	Hiring recommendations
POST	/chat/{profile_id}	Chat with profile
POST	/chat/clear	Clear session

---

Database Setup (MongoDB Atlas)

1. Create cluster at mongodb.com/cloud/atlas
2. Create database: docsmith
3. Create vector search index on resumes collection:

{
  "name": "vector_index",
  "type": "vectorSearch",
  "definition": {
    "fields": [{
      "type": "vector",
      "path": "embedding",
      "numDimensions": 768,
      "similarity": "cosine"
    }]
  }
}

4. Create standard indexes:

db.resumes.createIndex({ "extracted_data.name": 1 })
db.resumes.createIndex({ "extracted_data.skills": 1 })
db.resumes.createIndex({ "extracted_data.total_experience_years": 1 })
db.resumes.createIndex({ "status": 1 })
db.resumes.createIndex({ "created_at": -1 })

---

Project Structure

job-profile-screening/
├── backend/
│   ├── src/
│   │   ├── main.py              # FastAPI entry point
│   │   ├── agents/              # 8 AI agents (orchestrator, OCR, extraction, chat)
│   │   ├── api/routers/         # 5 REST routers (auth, resumes, jd, dashboard, chat)
│   │   ├── core/                # Config, logging, JWT security
│   │   ├── database/            # MongoDB connection, models, user repo
│   │   ├── memory/              # Chat history, shortlist sessions
│   │   ├── schemas/             # Pydantic models
│   │   ├── services/            # Groq, Gemini, MongoDB, search services
│   │   ├── utils/               # Image converter, text cleaner, prompts
│   │   └── workflows/           # Document processing pipeline
│   ├── data/                    # Uploads, page images, resumes, JDs
│   ├── Dockerfile
│   ├── requirements.txt
│   └── .env.example
├── frontend/
│   └── src/
│       ├── pages/               # 12 route-level pages
│       ├── components/          # UI (shadcn), auth, layout, search, upload, chat
│       ├── hooks/               # Custom React hooks
│       ├── lib/                 # API client, utilities
│       └── types/               # TypeScript interfaces
├── docker-compose.yml
├── ARCHITECTURE.md
└── README.md

---

Usage

1. Register at /auth as recruiter or job_seeker
2. Upload resumes at /upload (single or batch, PDF/image)
3. Review & verify extracted data on profile pages
4. Search candidates at /search using natural language queries
5. Upload JDs at /jd-search to auto-match candidates
6. Use AI assistants for shortlist refinement and hiring analysis
7. View analytics on the dashboard (skills, experience, domains)

---

Troubleshooting

Issue	Solution
ModuleNotFoundError	Activate venv, run pip install -r requirements.txt
MongoDB connection failed	Check MONGO_DB_URI and network access
Groq API errors	Verify API key and quota
PDF processing fails	Install Poppler
CORS errors	Backend allows http://localhost:8080 by default
Frontend blank page	Check browser console, verify backend is running

---

Related Docs

• Backend README - Setup and API details
• Frontend README - React setup
• ARCHITECTURE.md - System design

---

License

Proprietary software. All rights reserved.