Detect Hallucinations in Text with Custom Ollama MiniCheck

This n8n workflow provides automated fact-checking to detect hallucinations in AI-generated or user-provided text. It splits the input text into sentences, verifies each against a list of provided facts using the custom 'bespoke-minicheck' Ollama model, filters out inaccuracies, and generates a comprehensive summary with Qwen2.5, including the count of issues, listed problematic statements, and overall accuracy assessment.

Key benefits include rapid validation of content accuracy, reducing manual review time, and minimizing errors in high-stakes communications. It's particularly valuable for ensuring factual integrity in marketing materials, event descriptions, reports, and articles where misinformation can damage credibility.

Use cases span marketing teams validating event promos or ad copy, content creators checking blog posts, and researchers verifying summaries. Trigger manually or from other workflows, requiring only facts and text inputs after installing the Ollama model. Customizable for different fact sets or models, it focuses on verifiable statements while ignoring chit-chat.