Obligation Zero：以 Gemma 和 Ollama 分析文件並提取任務與期限

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady.

What I Built

Have you ever opened a long document just to find one important deadline buried somewhere in the text?

That is the problem I wanted to tackle with Obligation Zero — an AI-powered document analysis application that helps turn document content into actionable tasks.

Instead of manually scanning documents for every obligation, users can upload supported files and let the application identify important tasks, deadlines, and relevant evidence.

What can it do?

• 📄 Upload documents in PDF, DOCX, TXT, Markdown, CSV, and EML formats.
• 🔍 Extract obligations, tasks, and deadlines from document content.
• 🤖 Analyse document text using an Ollama-hosted AI model.
• ✍️ Generate draft replies to help users respond to obligations.
• ✅ Review extracted information and use it to plan their next steps.

The goal is simple: spend less time searching through documents and more time acting on what matters.

Demo

🌐 Live Application: https://obligation-zero.onrender.com/

🎥 Video Demo: https://youtu.be/51GEAb_vmmQ

Try the live application and see how a document can become a more actionable set of tasks.

Code

💻 GitHub Repository: https://github.com/rasikachavan13/Obligation-Zero.git

The project is built as a full-stack application, with a frontend, an Express backend, and AI-powered document analysis.

Note: Replace the GitHub profile link above with your exact Obligation Zero repository URL if the repository has a different name.

How I Built It

I built Obligation Zero as a full-stack web application, combining document processing, backend APIs, and open-source AI.

Technology stack

• Frontend: Vite-based web application
• Backend: Node.js and Express
• AI inference: Ollama Cloud
• AI model: Gemma, configured through Ollama
• Document processing: Text extraction for supported file formats
• Deployment: Render

How the workflow works

1. Upload: The user uploads a supported document.
2. Extract: The backend extracts available text from the document.
3. Analyse: The application sends the extracted content for AI-powered analysis.
4. Identify: The analysis identifies obligations, tasks, deadlines, and supporting evidence.
5. Draft: Users can generate draft replies to help them respond.
6. Review: Users review the results before deciding what action to take.

I also worked on fallback behaviour so that the application can provide rule-based suggestions when AI analysis fails. This is important because a useful application should handle failures instead of leaving users with a broken screen.

Scanned documents requiring OCR are not currently supported by the text-extraction workflow.

Why Does Open Innovation Matter?

Open innovation matters because it makes AI experimentation more accessible to developers who are still learning and building.

For Obligation Zero, using Gemma through Ollama gave me a way to explore AI-powered document analysis without making the entire application dependent on a proprietary, paid model API.

It also helped me understand more than just prompting a model. I had to work with backend integration, API authentication, document extraction, error handling, deployment, and the difference between a model being reachable and a model actually producing useful results.

Open tools make it easier to experiment, inspect workflows, adapt implementations, and learn by building. They help individual developers turn ideas into working applications without needing to build every component from scratch.

For me, open innovation is about having the freedom to experiment, learn, and build something useful while sharing that work with the developer community.

My Agent Session

I developed and debugged the application through an iterative workflow, testing document extraction, task and deadline identification, evidence references, saved draft replies, and persistence after refreshing the application.

I don't have a verified DevRelay agent-session link to share, so I'm leaving this optional section without an invented link.

The Lazy Payoff

The intended payoff is reducing the manual effort involved in reading lengthy documents and locating important obligations and deadlines.

Instead of searching through a document repeatedly, users can start with extracted tasks and deadlines, then review the relevant evidence.

I have built and deployed the application, but I haven't measured a reliable time-saving figure yet. I would like to test it with real users and compare the time spent reviewing documents manually against the time spent using Obligation Zero.

The goal isn't to remove human judgement. It's to reduce repetitive searching so people can focus on decisions and action.

Prize Categories

My submission is focused on the open-source AI theme, using Gemma through Ollama as part of the application.

I'll include any additional partner categories only where the project meets the challenge's eligibility requirements.

Thanks for checking out Obligation Zero! 🚀

If you have suggestions for improving document analysis, deadline extraction, or the overall workflow, I'd love to hear them in the comments.

Build steadily. Make useful things. Keep learning.