A Chrome extension that lets you select any online claim and instantly receive a grounded credibility assessment.

The Reality Button analyzes text (headlines, social posts, claims) and returns:

• Verdict (true / false / misleading / mixed / unclear)
• Confidence score
• Concise explanation
• Supporting sources

Misinformation spreads faster than verification. Most people don’t fact-check because it’s too slow and inconvenient.

This project reduces that friction to one click.

Frontend

• Chrome extension (popup UI + content script)
• Users can paste or select text directly on a webpage

Backend

• Node.js server deployed on Google Cloud Run

AI + Retrieval

• Uses Google Gemini (2.5 Flash)
• Enables live Google Search grounding
• Extracts real sources from search results
• Parses output into structured format

Architecture Chrome Extension → Cloud Run API → Gemini (with search grounding) → Structured response

The system uses live search instead of relying on model memory, which is critical for:

• current events
• political claims
• fast-changing information

Users provide their own Gemini API key:

• avoids centralized cost
• scales without usage limits
• mirrors real-world API products

Instead of forcing fragile JSON from the model:

• the system uses structured text
• parses deterministically on the backend
• improves reliability

Input "Vaccines cause autism"

Output

• Verdict: FALSE
• Confidence: 95%
• Explanation + sources (CDC, WHO)

cd Server
npm install
node server.js

Go to chrome://extensions Enable Developer Mode Click "Load unpacked" Select project folder

In the extension:

• Add your Gemini API key
• Set server URL (local or Cloud Run)

Working prototype. Actively improving:

• confidence calibration
• edge case handling
• UI/UX polish

better confidence scoring claim clustering / repeated detection caching to reduce API calls highlighting which sources support which claim