Rail-Madad (SIH 1711)

AI-driven autonomous workflow for railway complaint management with role-based dashboards, AI routing, and analytics.

Features

• AI complaint analysis and multi-department routing
• Role-based access (Passenger, Staff, Supervisor, Admin, Analyst)
• Complaint logging, status updates, and analytics dashboards
• Live chat assistant with train info lookup

Quickstart (Windows PowerShell)

python -m venv myenv
myenv\Scripts\activate
pip install -r requirements.txt

Run the full RBAC app (recommended)

streamlit run main_app.py

Run the minimal demo UI

streamlit run app.py

Configuration

The app reads secrets from Streamlit st.secrets and falls back to environment variables.

Required API keys

• GEMINI_API_KEY (Google Generative AI)
• RAPID_API_KEY (Indian Railway API via RapidAPI)
• SERPER_API_KEY (Google search via Serper)

Optional:

• GEMINI_MODEL_NAME (default: gemini-1.5-pro)

How to provide secrets

You can set them as environment variables before running Streamlit:

$env:GEMINI_API_KEY="<your_gemini_key>"
$env:RAPID_API_KEY="<your_rapidapi_key>"
$env:SERPER_API_KEY="<your_serper_key>"

Or create .streamlit/secrets.toml:

[api_keys]
GEMINI_API_KEY = "<your_gemini_key>"
RAPID_API_KEY = "<your_rapidapi_key>"
SERPER_API_KEY = "<your_serper_key>"

[model]
GEMINI_MODEL_NAME = "gemini-1.5-pro"

Gemini free key guide: Get API key

RapidAPI IRCTC playground: https://rapidapi.com/rahilkhan224/api/indian-railway-irctc/playground

Authentication & Roles

The RBAC app (main_app.py) loads auth/credentials.yaml. For local testing, copy and edit the example:

copy auth\credentials_example.yaml auth\credentials.yaml

All example accounts use password test123 by default.

Role	Username
Admin	admin1
Analyst	analyst1
Supervisor (Commercial)	supervisor_commercial
Staff (Commercial)	staff_commercial
Passenger	passenger1

You can customize users, roles, and departments in auth/credentials.yaml.

Data and Logs

• Complaints are stored in log_file.json at the repo root.
• When a complaint covers multiple departments, the app creates department-specific sub-complaints that share the same complaint number (cno).

Project Structure (high level)

• main_app.py: RBAC Streamlit application (multi-role dashboards)
• app.py: Minimal demo interface (Home, LiveChat, Directory, Lodger)
• agents.py, tasks.py, main.py: AI agents and Crew orchestration
• helperfunctions.py: complaint storage, analytics helpers
• tools.py: external tools (train status, search, wiki)
• auth/: RBAC helpers and credentials
• app_pages/: role-based page modules
• vars.py: department mapping and issue routing hints

Troubleshooting

• Missing Gemini key: AI features will be disabled and a Streamlit warning appears. Set GEMINI_API_KEY.
• Missing Serper key: Internet search tool returns a friendly message; chat still works without web search.
• Missing RapidAPI key: Train status tools return mocked responses.
• YAML not found: Create auth/credentials.yaml (copy from the example).

Gallery

Agent Workflow

Other Agent Architectures

Chatbot	Data Mining agent

Video Overview