Delhi Sahayak

Inspiration

As an aware citizen, I’ve seen firsthand that passion for change often dies at the feet of bureaucracy. Most citizens want to resolve issues like broken school infrastructure or poor sanitation, but they face a "Last Mile" information gap. General LLMs give vague advice, but they can't tell you the name of your specific Zonal Commissioner or how to escalate when a complaint is ignored. I was inspired to turn my frustration with this lack of escalation awareness into a tool that provides actual names, numbers, and accountability.

What it does

Delhi Sahayak is a hyper-local governance tool integrated into a community grievance portal. Unlike standard chatbots that offer general "how-to" guides, this app identifies the exact official responsible for a user's specific area and sector (e.g., Education in Rohini). It provides the official's name, verified contact info, and—most importantly—generates a pre-filled, formal escalation letter and provides actionable insights addressed to the next person in the hierarchy if the grievance remains unaddressed.

How we built it

We built a Hybrid RAG (Retrieval-Augmented Generation) architecture:

The Interface: A community portal designed for ease of use.

The Intelligence: An LLM-powered extraction layer that identifies "Department" and "Location" from raw user queries.

The Database: A structured data store (using a JSON/lookup-table approach for the POC) that maps Delhi’s administrative zones to specific personnel.

The Logic: A custom prompt engineering framework that prevents AI hallucinations by forcing the model to only use our verified directory for contact details.

Challenges we ran into

The primary challenge was administrative fragmentation. Delhi’s governance data is notoriously scattered across various department websites and PDFs. Mapping these complex hierarchies into a logic that an AI can query reliably was difficult. We also had to ensure that the LLM didn't just give a "helpful response" but strictly adhered to the structured data to ensure 100% accuracy in phone numbers and emails.

Accomplishments that we're proud of

We are incredibly proud of the Escalation Logic. Moving a citizen from "I have a problem" to "Here is a formal letter addressed to the Director of Education" in under 10 seconds is a massive win for civic empowerment. We successfully turned a "complaint box" into an "action engine."

What we learned

Building this taught us that information is the first step of activism. We learned how to bridge the gap between high-level AI capabilities and the grounded reality of civic bureaucracy. On the technical side, we realized that for high-stakes governance, a structured database used via "Tool Use" is far superior to a standard vector search that might hallucinate names.

What's next for Delhi Sahayak

The next phase is to automate the data ingestion by building scrapers for the Directorate of Education and MCD portals to keep the directory live. We also plan to add multi-language support (Hindi/Punjabi) and a "Tracking" feature where the community can collectively see which officials are responding to the escalation letters generated by the app.

Built With

• nextjs
• rag
• react
• vectordb