The problem InfoSauce solves

🎯 Problem
Information is fragmented across news websites, social media platforms and other online sources. Users need to switch between platforms, repeat searches, and compare conflicting claims to understand what is happening. This fragmentation creates several challenges:

1. Information overload: Too much content competes for attention.
2. Information gaps: Important developments may reach some people earlier than others.
3. Missed updates: Relevant information can disappear inside fast-moving feeds.
4. Misinformation: It is difficult to distinguish verified facts from rumors, opinions, or manipulated content.
5. Fear of missing out (FOMO): Users feel pressure to monitor multiple platforms continuously.
6. Time-consuming verification: Checking sources, dates, context, and competing claims requires significant effort.
7. Algorithmic filter bubbles: Platform feeds may prioritize engagement rather than relevance, accuracy, or completeness.

💡 Solution
InfoSauce is an AI-powered personal information radar that brings discovery, monitoring, summarization, analysis, and verification into one intelligent hub. It can gather relevant information from multiple sources, remove duplicates, highlight meaningful updates, and present the results in a clear and organized format. Instead of searching across numerous platforms, users can use InfoSauce to:

1. Discover the latest information in one place.
2. Track topics continuously and receive relevant updates.
3. Generate concise summaries of lengthy content.
4. Compare how different sources report the same event.
5. Trace important claims back to their original sources.
6. Identify conflicting information and missing context.
7. Reduce noise while avoiding important missed updates.

🤖 Gonka Router
InfoSauce uses Gonka Router as its AI-powered verification layer, sending extracted claims together with ranked evidence to multiple Gonka-hosted AI models for independent fact-checking. Their results are then compared through a multi-model consensus system, where disagreements are treated as uncertain, before InfoSauce calculates a Truth Score from 0–100 using model confidence, consensus, evidence coverage, and reliability. The system also returns the models' reasoning, supporting evidence, and Gonka Request IDs to provide a transparent and traceable verification process.

By combining multiple AI models, relevant external evidence, consensus-based verification, and confidence-aware scoring, InfoSauce aims to produce a more reliable and robust fact-checking result than relying on a single AI model alone.

Challenges we ran into

Technical Challenges Faced

1. Gonka Router API Integration — Implementing API authentication, request formatting, and communication with the Gonka inference gateway.
2. AI Model Reliability — Handling model failures such as HTTP 524 errors, malformed JSON, missing response content, and max_tokens termination.
3. Multi-Model Consensus — Designing a system that sends the same claim to multiple AI models and reliably handles agreement, disagreement, and failed models.
4. Truth Score Algorithm — Designing and implementing a scoring system that combines model confidence, consensus, evidence coverage, and reliability into a 0–100 score.
5. Evidence Retrieval & Ranking — Retrieving relevant evidence from external sources, filtering unusable results, and ranking the strongest evidence before sending it to the AI models.
6. Claim Extraction — Automatically extracting the core factual claim from articles and user-submitted content before verification.
7. Multi-Platform Data Integration — Handling different data structures and APIs from Reddit, X/Twitter, YouTube, Bilibili, Xiaohongshu, and web search, then normalizing them into a common format.
8. URL Content Extraction — Building platform-specific URL detection and content extraction, including dealing with Reddit's post structure and identifiers.
9. Backend–Data Service Communication — Connecting the Express backend to the separate Data service and validating responses between the two services.
10. Production Deployment Constraints — OpenCLI and browser-based integrations worked locally but some failed on Render because the required Browser Bridge/Chrome environment was unavailable.
11. Performance & Concurrency — Managing the high latency of retrieving evidence and running multiple AI models while preventing too many operations from running simultaneously.
12. Error Handling & Validation — Building safeguards throughout the pipeline to prevent invalid input, bad API responses, missing evidence, failed models, and malformed verification results from breaking the system.