README.md

metadata

title: Gov Env - Government Service Assistant
colorFrom: blue
colorTo: green
sdk: docker
app_file: server/app.py
pinned: false

🇮🇳 Government Service Application Assistant Environment

A production-quality OpenEnv environment that simulates an AI assistant helping users navigate Indian government service applications, with a focus on document validation and correction workflows.

🎯 Problem Statement & Motivation

The Challenge

Every year, ~1.3-1.4 crore (13-14 million) Indian citizens apply for passports alone. With 138+ crore Aadhaar holders and 2.5-3 crore PAN applications, the scale is massive.

The Pain Point

70-80% of government service rejections are due to:

• Document confusion and missing requirements
• Name mismatches across documents
• Invalid or expired proof documents

Our Solution

An AI agent environment that trains models to:

1. Identify required documents for any government service
2. Validate submitted documents for completeness and validity
3. Suggest corrections for rejected applications

---

📋 Action Space

{
  "action_type": "validate_documents",
  "service_type": "passport_new",
  "documents": [{"type": "Aadhaar Card", "details": "Valid"}]
}

Action	Description
select_service	Choose a government service
list_required_documents	Get document requirements
validate_documents	Check submitted documents
suggest_corrections	Get fix suggestions
submit_application	Finalize application

---

📝 Task Descriptions

Task ID	Difficulty	Description	Optimal Steps
passport_new_easy	Easy	Identify required documents	2
passport_new_medium	Medium	Validate submitted docs	3
passport_new_hard	Hard	Fix incorrect application	4

---

💰 Grading System

• Range: [0.0, 1.0]
• Deterministic: Same input → same output
• Components: Document accuracy (0-0.3) + Validation quality (0-0.3) + Efficiency (+/- 0.1)

---

🚀 Endpoints

Endpoint	Method	Description
/reset	POST	Initialize environment
/step	POST	Execute action
/state	GET	Get current state
/tasks	GET	Get available tasks
/health	GET	Health check

---

🤖 Running Inference

export HF_TOKEN=your_groq_api_key
export ENV_URL=https://dharunkkk-gov-env.hf.space
python3 inference.py passport_new_easy

Output Format

[START] task=passport_new_easy env=gov_env model=llama-3.3-70b-versatile
[STEP] step=1 action=... reward=1.00 done=false
[END] success=true steps=10 score=10.00

---

✅ Validation Checklist

• HF Space deploys (HTTP 200)
• OpenEnv spec compliance
• Dockerfile builds
• Baseline runs without errors
• 3+ tasks with graders
• Grader outputs in [0.0, 1.0]

---

📁 Project Structure

├── inference.py
├── Dockerfile
├── requirements.txt
├── openenv.yaml
├── models.py
├── tasks.py
├── grader.py
├── client.py
└── server/
    ├── app.py
    └── gov_env_environment.py

---

Live Space: https://dharunkkk-gov-env.hf.space