Red Sum
TYPE
mobile
DURATION
2 Days
COMPLETED
July 2025
LANGUAGE
Dart
MISSION BRIEF
Red Sum is an intelligent Reddit thread summarization and analysis app built with Flutter. It leverages Firebase AI, Supabase, and the Gemini Live API to generate high-quality summaries of Reddit posts, extract diverse viewpoints from comment sections, and provide users with multiple ways to interact with the content. Users can listen to AI-generated 90–120 second podcasts summarizing the post, chat with the AI to explore deeper insights, or even engage in live voice conversations about the post’s context. Red Sum brings clarity, accessibility, and engagement to the massive world of Reddit discussions.
VISUAL_EVIDENCE
CORE FEATURES
- Automatic Reddit post summarization
- Comment viewpoint extraction and clustering
- AI-generated 90–120 second podcast summaries
- Interactive AI chat about the post
- Live voice-to-voice conversation with AI
- Save and bookmark Reddit posts
- Share summaries and podcasts externally
- Offline access for saved posts
- Dark/light theme modes
- Multi-language summarization support
TECH STACK
- >Flutter for cross-platform UI (Android & iOS)
- >Firebase AI for summarization and viewpoint extraction
- >Supabase for backend data storage and authentication
- >Gemini Live API for context-aware conversation and podcast generation
- >Real-time transcription and TTS for voice interactions
- >Vector embeddings for comment clustering and similarity ranking
- >Cached summaries for offline use
- >Secure REST and WebSocket communication layers
- >Material 3 adaptive design
Operational Hurdles
Accurately extracting distinct viewpoints from thousands of Reddit comments
SOLUTION_PROTOCOL: Used semantic clustering with vector embeddings and sentiment tagging to group comments into coherent viewpoints
Maintaining coherent AI conversation context across long Reddit threads
SOLUTION_PROTOCOL: Implemented a hierarchical summarization approach and persistent chat memory within Gemini Live sessions
Generating natural and engaging podcast narration
SOLUTION_PROTOCOL: Used multi-voice synthesis models and dynamic emphasis tagging for natural speech generation
Ensuring smooth real-time voice interaction with minimal latency
SOLUTION_PROTOCOL: Leveraged Gemini Live streaming APIs with audio buffering and WebSocket-based voice exchange