Sentinelle
A credibility assessment layer for short video content in English, Chinese, Malay, and Tamil, with supporting details available directly in the feed.
Hackomania · 7 March 2026
Problem and Context
Short videos make claims quickly, often without giving viewers time to inspect their sources. Sentinelle explores bringing claim analysis into the viewing experience across Singapore's four official languages.
Analysis Flow
- Upload a video to Firebase Storage.
- Transcribe its content with Gemini's multimodal analysis.
- Extract claims into structured fields.
- Use web search grounding to assess the claims and generate a credibility score.
- Open a feed badge to inspect the assessment and supporting details.
Architecture
React Native, Expo, TypeScript, and Redux Toolkit provide the mobile interface. Firebase AI Logic connects it to Gemini 2.5 Flash for transcription, claim extraction, and grounded analysis. Firebase Storage holds video uploads, and Firestore stores posts and results.
The prototype includes sample videos and an anonymous demo mode.
Project Credits
Built by the Code and Conquer team for Hackomania.