This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
What I Built
Scope creep is one of those problems freelancers know too well.
A project starts with a clear agreement, but then the client asks for one more feature, another revision, or an extra deliverable. These requests can slowly turn a profitable project into unpaid work.
That's why I'm building ScopeSentry, a privacy-first AI tool that helps freelancers compare their original project agreement with client conversations.
You paste both texts, and ScopeSentry analyzes the requests and categorizes its findings into four verdicts:
-
in_scope— the request appears to be covered by the agreement -
potentially_out_of_scope— the request may go beyond the agreed work -
unclear— the agreement doesn't provide enough clarity -
needs_review— the finding needs further verification
Each finding includes supporting text from the original sources, so you can review the evidence instead of blindly trusting an AI response.
ScopeSentry is designed to help freelancers make better decisions before agreeing to extra work.
Code
GitHub repository: https://github.com/sufiankhan-dev/ScopeSentry
The project is being built in the open. You can explore the code, report issues, and contribute improvements.
How I Built It
I'm building ScopeSentry with open-source AI at its core.
The main stack includes:
- Next.js and TypeScript for the application
- Gemma as the language model
- Ollama to run the model locally
- Zod for validating structured data
- Deterministic checks to verify that evidence quotes actually appear in the supplied source text
The analysis compares the project agreement with the client's requests and produces structured findings.
One important design decision is that missing information in an agreement shouldn't automatically mean a request is out of scope. ScopeSentry should recognize ambiguity and flag uncertain findings for human review.
Privacy is central to the design. The goal is to run inference locally without sending project agreements or client conversations to a cloud AI provider.
Why Does Open Innovation Matter?
Freelancers often work with contracts, client messages, budgets, and other sensitive information. Sending all of that to an external AI service isn't always desirable.
Open-weight models such as Gemma make it possible to build tools that people can run on their own machines, giving them more control over their data and how AI is used.
Open innovation also lets developers inspect the implementation, improve the evidence checks, experiment with different models, and contribute features without depending entirely on a closed AI API.
For ScopeSentry, that control is a core part of the product rather than just a technical choice.
My First User
I'm building ScopeSentry around a problem I want to validate with real freelancers: how often do extra client requests fall outside the original agreement, and what evidence would help freelancers respond confidently?
I'm looking for my first users to test it with realistic project briefs and client requests, identify incorrect or unclear verdicts, and share feedback on whether the findings would actually help them handle scope changes.
If you're a freelancer or independent developer and have dealt with scope creep, I'd appreciate your feedback. I don't have a confirmed user testimonial or paid user to report yet, so this is an invitation to help test and improve the project rather than a claim of existing traction.
Prize Categories
- Gemma — Best Use of Gemma: ScopeSentry uses Google's open-weight Gemma model through Ollama for local analysis.
If you've ever had a client ask for extra work that wasn't part of the original agreement, I'd love to hear how you handled it and what you'd want a tool like this to do.
Top comments (0)