Fact Pulse

A Flutter application that uses Perplexity’s Sonar API to provide real-time fact-checking for images and spoken content. Fact Pulse helps users verify information on-the-fly during debates, presentations, or when analyzing images containing claims.

Perplexity Hackathon Submission

Category: Information Tools
Bonus Category: Deep Research & Education

Demo & Screenshots

Watch the demo video

Web Demo - May remove later

Features

• Real-time Speech Analysis Transcribes spoken content and fact-checks statements on the fly.
• Image Fact Verification Analyzes images and provides fact-checking for visual content.
• Claim Rating System Categorizes claims as TRUE, FALSE, MISLEADING, or UNVERIFIABLE.
• Source Citations Provides reliable sources for each fact-check.
• User Authentication Secure login with Firebase (Google, Email).
• Debate Mode Continuous speech recognition during live debates, with instant fact feedback.
• Firestore Integration Persists transcripts and fact-checks for future reference.
• Cross-platform Supports iOS, Android, and Web (macOS support in testing, Windows not currently supported).

Perplexity API Integration

Fact Pulse leverages Perplexity's Sonar API in several key ways:

1. Debate Fact-Checking:

   • Speech is transcribed in real-time using speech_to_text
   • Transcribed text is sent to Perplexity's Sonar API via our custom client
   • Carefully crafted prompts (in lib/debate/prompts.dart) instruct the model to:
     • Extract verifiable factual claims
     • Verify each claim against authoritative sources
     • Return structured JSON with claim, rating, explanation, and sources

2. Image Analysis:

   • Images are processed and sent to Sonar API
   • The API extracts text from images and identifies claims
   • Each claim is verified and rated for accuracy

3. JSON Response Handling:

   {
     "claims": [
       {
         "claim": "Example factual statement",
         "rating": "TRUE|FALSE|MISLEADING|UNVERIFIABLE",
         "explanation": "Verification details",
         "sources": ["Source URL 1", "Source URL 2"]
       }
     ]
   }

Open-Source Packages

As part of this project, I've developed and open-sourced two Dart packages to make Perplexity API integration easier for the Flutter community:

perplexity_dart

A lightweight, type-safe Dart SDK for Perplexity's chat/completions API with both streaming and non-streaming support.

• pub.dev
• GitHub

Key features:

• Type-safe API client
• Streaming and non-streaming support
• Comprehensive error handling
• Customizable request options

perplexity_flutter

Flutter-specific wrapper and widgets for easier Perplexity API integration in Flutter apps.

• pub.dev
• GitHub

Key features:

• Ready-to-use Flutter widgets
• Simplified state management
• Platform-specific optimizations
• Example implementations

Technical Implementation

Architecture & State Management

Built with Flutter and the BLoC pattern for a clean separation between UI and logic.

• Authentication Firebase Auth (Google, Apple, Email)
• Debate Analysis speech_to_text for streaming transcription + Perplexity Sonar API for fact checks
• Image Analysis Upload or URL-reference images to Sonar API, extract and verify on-image text
• Data Layer Firestore collections per user → debates/images → transcriptions & claims

Firestore Schema

users/{uid}/debates/{debateId}/
  • claims (sub-collection)
  • { topic, createdAt, response, complete, updatedAt }

users/{uid}/speechs/{speechId}/
  • claims (sub-collection)
  • { topic, createdAt, response, complete, updatedAt }

users/{uid}/images/{imageId}/
  • claims (sub-collection)
  • { topic, createdAt, response, complete, lastOpenedAt }

Impact & Innovation

Fact Pulse addresses the critical challenge of misinformation by:

1. Democratizing Fact-Checking: Puts powerful verification tools in everyone's hands
2. Real-Time Analysis: Provides immediate feedback during live conversations
3. Educational Value: Helps users learn to identify reliable information
4. Open Source Contribution: The packages we've developed make Perplexity's powerful API accessible to the entire Flutter ecosystem

Current Limitations

• Apple Sign-in: Not currently implemented as it requires a paid Apple Developer membership
• Firebase Storage: Image and report storage is limited as Firebase Storage requires a paid subscription for larger storage needs
• API Models: Currently using only Sonar and SonarPro models due to limited Perplexity API credits
• Platform Support: Windows is not currently supported, and macOS support is still in testing

These limitations are primarily due to development resource constraints and would be addressed in a production environment.

Setup & Run

1. Clone

   git clone https://github.com/vishnu32510/fact_pulse.git
   cd fact_pulse

2. Dependencies

   flutter pub get

3. Firebase

   • Create a Firebase project
   • Enable Auth (Google, Email) & Firestore
   • Use Firebase CLI to configure local Firebase settings

4. API Key

   • Set PERPLEXITY_API_KEY in your environment or .env
   • Set PERPLEXITY_API_KEY in lib/perplexity.dart (not recommended for production)

5. Run

   flutter run

Future Roadmap

• Real-time collaborative sessions
• News article verification on date range
• Video-conference integration
• Offline/cached fact-checks
• Misinformation analytics dashboard
• Browser extension for web content verification

License

MIT License — see the LICENSE file for details.