CoolFace
Apppublic

Vraj07/Scheme-Advisor-Agentic-Environment

sourceHugging Faceupdated 6mo agoView on Hugging Face
0likes
App README

Scheme Advisor Agentic Environment

Overview

Scheme Advisor is a high-fidelity agentic environment designed for AI agents to learn the task of Government Welfare Scheme Counseling. In this environment, an agent takes on the role of a social worker or CSC (Common Service Centre) operator in India, helping citizens identify which government welfare schemes they are eligible for and what documents they need to collect.

How it Works

The environment is structured as a multi-step reinforcement learning episode (open-loop or closed-loop).

  1. 1.Context: Each episode starts with a Citizen Context (a narrative description of a citizen's life, occupation, income, and needs).
  2. 2.Profiling: The agent must intelligently extract and submit the citizen's profile using submit_profile.
  3. 3.Discovery: Based on the profile, the environment suggests potentially eligible schemes.
  4. 4.Inquiry: The agent queries specific schemes (query_scheme) to understand their benefits and document requirements.
  5. 5.Advice: The agent concludes the session by requesting the final set of required documents (request_documents).

Key Components

  • —Knowledge Base (`schemes_db.py`): A structured database of major Indian welfare schemes (PM-KISAN, Ayushman Bharat, etc.) with deterministic eligibility rules.
  • —Environment Engine (`server/environment.py`): Handles the state transitions, action processing, and rewards.
  • —Scoring System: Graded reward based on:
  • —Profile Accuracy (30%): Correctly extracting information from the narrative.
  • —Scheme Recall (30%): Finding all truly eligible schemes.
  • —Scheme Precision (20%): Avoiding "hallucinating" eligibility for schemes the citizen doesn't qualify for.
  • —Document Accuracy (20%): Specifying the exact documents needed.

Getting Started

Prerequisites

  • —Python 3.10+
  • —FastAPI & Uvicorn

Installation

bash
pip install -r server/requirements.txt

Running the Server

bash
uvicorn server.app:app --host 0.0.0.0 --port 7860

Running the Baseline Agent

You can test the environment using the built-in baseline script:

bash
python baseline.py

Documentation

  • —API_GUIDE.md: Detailed explanation of all API endpoints and CURL commands.
  • —TESTING_GUIDE.md: Step-by-step instructions for manual and automated testing.

Technology Stack

  • —Backend: FastAPI (Python)
  • —Architecture: OpenEnv (Agentic Environment Standard)
  • —Models: Pydantic v2
  • —Deployment: Hugging Face Spaces Docker Runtime