-
-
Start with the message. Paste an important written message to begin the analysis.
-
A clearer picture of the message. Review the findings organized into four categories.
-
See the supporting evidence. Inspect the original passage behind a finding.
-
Spot missing details. Identify relevant information that the supplied text doesn't provide.
-
Identify what's unclear. See where wording or conditions leave room for interpretation.
-
Know what to clarify. Review follow-up questions that may help resolve uncertainty.
-
Review the original contents.
Inspiration
Important messages, such as job offers, emails, and official announcements, can contain unclear conditions or leave out details that people need to understand. I wanted to build something that could help people examine these messages more carefully instead of relying on a quick first impression.
That idea led to SignalCheck, a tool designed to help people understand what a message actually says and what they may need to clarify.
What It Does
SignalCheck allows users to paste an important written message and analyze it using four categories:
-Facts: What the supplied text explicitly states.
-Ambiguities: What is unclear or open to interpretation.
-Information Not Stated: Relevant details that the message does not provide.
-Questions to Ask: Follow-up questions that may help clarify the situation.
The app can also show the original passage supporting a finding, allowing users to compare the analysis with the text they provided. SignalCheck is designed to support careful review, not to make decisions on behalf of users.
How I Built It
I built SignalCheck using React and Vite for the frontend, with Node.js and Express handling the backend. I integrated Google's Gemini API to generate structured analyses and used Zod to validate the expected response format.
I also implemented input-length limits and evidence validation to help ensure that findings presented as facts or ambiguities are supported by the supplied text. I deployed the application to Netlify so people can try it online.
Challenges and What I Learned
One of the major challenges I faced was making the analysis useful without treating every AI-generated statement as a fact. I also had to pay particular attention to checking evidence against the original message and keeping the different result categories distinct.
Deploying the backend as a serverless function was another part of getting the application working online. Through this project, I learned much more about integrating an AI API into a web application, validating structured AI output, handling deployment configuration, and testing the complete user experience.
Built With
- codex
- express.js
- geminiapi
- javascript
- netlify
- node.js
- react
- vite
- zod