🌟 Inspiration I’ve always been fascinated by how government schemes can transform lives—but only if people can access and understand them. The idea for The OpenGov Orbit came from a simple question: What if we could make policy documents as searchable and intuitive as Google, even offline? That spark led me to explore open-weight models and build a tool that could answer questions about government schemes using only official documents.
🚀 What it does The OpenGov Orbit is a lightweight, offline Q&A tool that uses a locally deployed language model to answer questions based solely on the National Food Security Act, 2013. It reads a local copy of the act and responds to user queries with expert-level precision—without relying on internet access or external databases. It’s fast, private, and focused.
🛠️ How I built it
Tech Stack: Python, requests library, and Ollama serving the gpt-oss-20b model.
Data Source: Key articles from the National Food Security Act saved in scheme.txt.
Core Logic: My script reads the document and constructs a prompt that instructs the model to answer only using the provided text. This prompt is sent to the local API.
Interface: A clean command-line interface that showcases the model’s capabilities without distractions.
⏱️ Challenges I ran into Time was the biggest constraint. I had to make tough decisions quickly—like choosing Ollama over a manual Hugging Face setup to avoid configuration delays. Another challenge was prompt guarding: early versions of the prompt allowed the model to pull in general knowledge. I had to iterate multiple times to ensure it would only respond based on the source document, or refuse to answer otherwise.
🏆 Accomplishments that I'm proud of Deploying a 20B parameter model locally with a single command.
Building a functional prototype in under five hours.
Achieving reliable, document-bound responses through prompt engineering.
Proving that powerful AI tools can run offline on consumer hardware.
📚 What I learned Open-weight models have matured dramatically—they’re now accessible, fast, and surprisingly capable.
Prompt engineering is an art form. The way instructions are phrased can make or break the reliability of responses.
Simplicity in interface design helps highlight core functionality, especially in time-constrained builds.
đź”® What's next for The OpenGov Orbit I want to expand the tool to cover multiple government acts and schemes, creating a modular system where users can plug in different documents. A web-based interface, multilingual support, and voice input are also on the roadmap. Ultimately, I envision The OpenGov Orbit as a trusted offline companion for citizens, activists, and policy researchers alike.
Log in or sign up for Devpost to join the conversation.