Swavalambi — AI-Powered Platform for Skilled Workers

Swavalambi helps Indian artisans and skilled workers find jobs, access government schemes/loans, and validate their skills using AI.

Built for the AWS AI for Bharat Hackathon by Team Algonauts.

---

Features

• 🤖 AI Profiling Chat — Conversational agent (Claude via Bedrock/Anthropic) collects skill, intent, experience, and preferred job location
• 📸 Vision Skill Assessment — Upload a work photo; AI scores it and saves a skill_rating to your profile
• 💼 Personalised Job Search — Filtered by skill + preferred city/state via NCS (National Career Service) API
• 📚 Training Centers — Matched by skill using live Skill India API with local JSON fallback
• 🏛️ Government Schemes — Searched via myScheme.gov.in API
• 🔐 Dual Authentication — Email/password (Cognito) or OTP (mock, for testing)
• 🌐 Voice Support — Hindi voice input/output with AWS Transcribe / Polly / Sarvam AI
• 🎯 Personalized Greeting — AI addresses user by name and skips asking for it if already known
• 🔒 Skill-gated Access — Application buttons locked below Level 3 skill rating

---

Project Structure

swavalmbi-ai/
├── backend/
│   ├── agents/
│   │   ├── profiling_agent.py     # AI chat — collects skill, intent, location
│   │   └── vision_agent.py        # Score work sample photos
│   ├── api/
│   │   ├── routes_auth.py         # Register, login, OTP, Cognito
│   │   ├── routes_chat.py         # POST /chat/chat-profile
│   │   ├── routes_users.py        # User CRUD + delete from DynamoDB & Cognito
│   │   ├── routes_recommendations.py  # Jobs, schemes, training centers
│   │   ├── routes_vision.py       # Photo skill assessment
│   │   └── routes_voice.py        # Voice transcription / synthesis
│   ├── services/
│   │   ├── dynamodb_service.py    # DynamoDB CRUD helpers
│   │   ├── cognito_service.py     # Cognito register / login / admin delete
│   │   └── live_data.py           # NCS jobs, myScheme, SkillIndia APIs
│   ├── schemas/
│   │   └── models.py              # Pydantic request/response models
│   ├── data/
│   │   └── upskill-agent/
│   │       └── skill_india_training_centers.json  # Local fallback (~25MB)
│   └── scripts/
│       ├── setup_dynamodb.py
│       ├── setup_cognito.py
│       └── setup_voice_services.py
└── frontend/
    └── src/pages/
        ├── Login.tsx       # OTP + email/password auth
        ├── Register.tsx    # Email/password registration
        ├── Home.tsx        # Dashboard + recommendations
        ├── Assistant.tsx   # AI chat interface
        ├── Upskill.tsx     # Training center listings
        ├── Schemes.tsx     # Government scheme listings
        └── Profile.tsx     # User profile + skill level

---

Prerequisites

• Python 3.10+ (conda env: ai4bharat)
• Node.js 18+ & npm
• AWS credentials with access to:
  • Amazon Bedrock (Claude Sonnet — ProfilingAgent + VisionAgent)
  • Amazon DynamoDB (us-east-1)
  • Amazon Cognito (User Pools — email/password auth)
  • Amazon S3 (voice audio files)
  • Amazon Transcribe, Polly, Translate (voice features)

---

First-Time Setup

1. Create AWS Resources

cd backend
conda activate ai4bharat

python scripts/setup_dynamodb.py      # Creates swavalambi_users table
python scripts/setup_cognito.py       # Creates Cognito User Pool + Client
python scripts/setup_voice_services.py  # Creates S3 bucket

⚠️ If your AWS session token has expired, run aws sso login first.

2. Configure backend/.env

# AWS
AWS_ACCESS_KEY_ID=...
AWS_SECRET_ACCESS_KEY=...
AWS_SESSION_TOKEN=...        # optional — only if using SSO/temporary credentials
AWS_DEFAULT_REGION=us-east-1

# DynamoDB
DYNAMODB_TABLE=swavalambi_users

# Cognito (from setup_cognito.py output)
COGNITO_USER_POOL_ID=us-east-1_XXXXXXXXX
COGNITO_CLIENT_ID=xxxxxxxxxxxxxxxxxxxxxxxxxx

# LLM — pick one
USE_ANTHROPIC=true
ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_MODEL_ID=claude-3-5-sonnet-latest

# or Bedrock
USE_ANTHROPIC=false
BEDROCK_MODEL_ID=global.anthropic.claude-sonnet-4-5-20250929-v1:0

# Voice (optional)
VOICE_PROVIDER=aws
AWS_S3_BUCKET=swavalambi-voice
AWS_POLLY_VOICE_ID=Aditi

---

Running Locally

Backend

cd backend
conda activate ai4bharat
pip install -r requirements.txt   # first time only
python -m uvicorn main:app --reload --port 8000

Swagger UI → http://localhost:8000/docs

Frontend

cd frontend
npm install       # first time only
npm run dev

App → http://localhost:3000

---

API Endpoints

Authentication

Method	Path	Description
POST	/api/auth/register	Register with email + password → verification email sent
POST	/api/auth/verify-email	Confirm email with 6-digit code
POST	/api/auth/login	Login with email + password → JWT tokens
POST	/api/auth/resend-code	Resend verification code
POST	/api/auth/forgot-password	Start password reset
POST	/api/auth/reset-password	Confirm reset with code
POST	/api/auth/send-otp	Legacy OTP send (mock 123456, testing only)
POST	/api/auth/verify-otp	Legacy OTP verify (testing only)

User Management

Method	Path	Description
POST	/api/users/register	Upsert user profile in DynamoDB
GET	/api/users/{user_id}	Fetch user profile
DELETE	/api/users/{user_id}?email=...	Delete from DynamoDB + Cognito
GET	/api/users/{user_id}/chat-history	Fetch chat history
DELETE	/api/users/{user_id}/chat-history	Clear chat history

AI Services

Method	Path	Description
POST	/api/chat/chat-profile	Profiling conversation — collects skill, intent, location
POST	/api/vision/analyze-vision	Score work sample photo
POST	/api/recommendations/fetch	Jobs / schemes / training centers (location-filtered)

Voice Services

Method	Path	Description
POST	/api/voice/transcribe	Audio → text (Transcribe / Sarvam)
POST	/api/voice/synthesize	Text → speech (Polly / Sarvam)
POST	/api/voice/translate	Translate text

Health

Method	Path	Description
GET	/health	Health check

---

User Flow

Register (email + password)  →  Verify email
    ↓
Login  →  Dashboard
    ↓
AI Chat Assistant:
  1. Greets by name (already known from DB)
  2. Asks profession/skill
  3. Asks intent: Job / Upskill / Loan
  4. If Job → asks preferred city/state
  5. Assesses experience (beginner / intermediate / advanced)
  6. If intermediate/advanced → asks to upload work photo
    ↓
Vision Agent scores photo  →  skill_rating saved to DynamoDB
    ↓
Dashboard shows personalised:
  - Jobs (filtered by skill + location via NCS)
  - Government Schemes (via myScheme.gov.in)
  - Training Centers (live Skill India API + local JSON fallback)

Skill Lock Logic

Skill Rating	Access
< 3 (Beginner)	View-only — apply buttons locked
≥ 3 (Intermediate/Advanced)	Full access — can apply for jobs, loans, schemes

---

Deleting a Test User

To fully reset a user (remove from both DynamoDB and Cognito):

# Via curl
curl -X DELETE "http://localhost:8000/api/users/9898989898?email=test@example.com"

# Or open Swagger UI → DELETE /api/users/{user_id}

---

Recommendation Data Sources

Data Type	Source	Fallback
Jobs	NCS API (filtered by skill + location)	None
Schemes	myScheme.gov.in API	None
Training Centers	Skill India Digital Hub API	Local JSON (data/upskill-agent/)

The local JSON fallback is used when the live Skill India API returns no skill-matched centers (e.g. for "Plumber", the JSON has "Plumber - General" courses).

---

Tech Stack

Layer	Technology
Frontend	React 18, Vite, TypeScript, Vanilla CSS
Backend	FastAPI, Python 3.10+, Uvicorn
AI Agents	AWS Bedrock / Anthropic Claude (Strands Agents SDK)
Authentication	AWS Cognito (User Pools, JWT)
Database	Amazon DynamoDB (swavalambi_users)
Voice	AWS Transcribe + Polly + Translate / Sarvam AI
Storage	Amazon S3
Jobs API	NCS — National Career Service
Schemes API	myScheme.gov.in
Training API	Skill India Digital Hub + local JSON fallback