Clause terminator

To read files and documents to identify hidden clauses or loopholes which could completely make the agreement or deal rigged or one sided.

Comment 2

Inspiration

Terms of Service agreements, rental leases, and employment contracts are intentionally designed to be skipped. They are long, dense, and written in exhausting legalese. The average person signs away their privacy, data, and legal rights every single day simply because they don't have the time or a law degree to read between the lines. We wanted to build a digital shield something that instantly uncovers what you are actually agreeing to. That is why we built Clause Terminator: an elite, AI-powered document auditor that shreds through legalese to protect everyday users.

What it does

Within seconds, Clause Terminator strips away the filler and provides a clean, visual dashboard:

-->Risk Assessment: Instantly gauges the document's safety profile (Low, Medium, High, or Critical).

-->Red Flags & Gotchas: Exposes aggressive auto-renewals, hidden fees, forced arbitration, or excessive data privacy waivers.

-->User Obligations: Extracts exactly what you are required to pay, do, or forfeit.

-->Escape Routes: Highlights precisely how you can legally terminate the agreement and what the penalties are.

If a research paper is uploaded instead, it pivots seamlessly to highlight the methodology, logical leaps, and the actual bottom line.

How we built it

-->Frontend & Dashboard: Built entirely in Python using Streamlit. This allowed us to build a responsive, highly visual user interface with interactive file uploaders and color-coded warning metrics without touching a line of HTML or CSS.

-->The AI Engine: We created a drop down list to use any gemini model that we require for any docuement that we need maximizing efficiency and precision when needed.

-->Strict JSON Extraction: We configured Gemini with low temperature 0 and utilized the Structured Outputs response schema feature in Google AI Studio. This forced the model to return deterministic, perfectly structured JSON, allowing our Python backend to reliably map data directly to the user interface.

Challenges we ran into

Our biggest hurdle was preventing the AI from generating conversational filler Here is your analysis which instantly crashes code that expects pure JSON. We solved this by diving deep into the system instructions and enforcing strict schema validation directly at the API level.

Another challenge was ensuring the model could correctly distinguish between a legal contract and an academic research paper on the fly, requiring careful prompt engineering to build a dynamic analytical pipeline.

Accomplishments that we're proud of

We are incredibly proud of building a fully functional, end-to-end MVP within the hackathon timeline. Getting the app to parse complex text, send it to an LLM, and render perfectly styled "Red Flag" warning cards dynamically based on the AI's risk assessment felt like magic when it finally ran perfectly.

What we learned

We learned the immense value of low-temperature modeling and structural constraints when building user-facing apps. We also discovered just how fast you can go from an abstract idea to a working, beautiful prototype when combining the prototyping speed of Streamlit with the raw power of the Gemini API.

What's next for Clause terminator

Clause Terminator is just getting started. Our next steps include:

-->OCR Integration: Allowing users to take a live photo of a paper document with their phone and audit it on the spot.

-->Browser Extension: A companion extension that analyzes Terms of Service pages automatically as you browse the web, warning you before you click "I Agree."

-->Crowdsourced Risk Database: Building an open-source ledger where users can contribute audited versions of common tech platforms' terms of service.

Built With

Share this project:

Updates