AI Fact-Checking & Error Summarization Agent

Unlock Automated Fact-Checking and Error Reporting with this AI Agent

This powerful AI Agent automates the critical task of content verification. It takes a source document (the 'facts') and a text to be checked, intelligently splitting the text into individual claims. Each claim is then meticulously compared against the source document using a specialized local Ollama LLM (bespoke-minicheck) to determine its accuracy (Yes/No). Finally, another local Ollama LLM (qwen2.5:1.5b or your configured model) synthesizes a structured summary detailing all identified incorrect statements, ready for your review and action. This agent empowers you to maintain high content quality and accuracy with minimal manual effort, all while keeping your data private on your local infrastructure.

Key Features & Benefits

• AI-Powered Claim Verification: Leverages a specialized bespoke-minicheck Ollama model for precise "Yes/No" fact-checking of individual sentences/claims against a provided source document.
• Intelligent Error Summarization: Employs a qwen2.5:1.5b Ollama model (or your configured alternative) to generate a clear, structured summary of all incorrect statements, highlighting what needs correction based on a detailed system prompt.
• Local & Private Processing: Operates entirely with local Ollama models, ensuring your sensitive data remains on your infrastructure, offering enhanced privacy and control.
• Automated Content Splitting: Intelligently breaks down the input text into sentences for granular analysis, preserving dates and list structures.
• Efficient Workflow: Streamlines the otherwise time-consuming process of manual fact-checking and error compilation.
• Actionable Output: Provides a direct list of inaccuracies and an overall assessment, making it easy for content teams to make revisions.
• Flexible Integration: Can be triggered manually for ad-hoc checks with sample data or integrated into larger content pipelines by other n8n workflows, accepting 'facts' (source document) and 'text' (document to check) as inputs.

• An n8n instance (Cloud or self-hosted).
• Ollama installed and running locally, accessible by n8n.
• The bespoke-minicheck:latest Ollama model pulled (run ollama pull bespoke-minicheck:latest).
• The qwen2.5:1.5b Ollama model pulled (run ollama pull qwen2.5:1.5b), or an alternative summarization model configured in the workflow.
• Configured n8n credentials for your Ollama API.

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Ensure your Ollama instance is running and accessible by n8n, and that the required models (bespoke-minicheck:latest, qwen2.5:1.5b) are pulled.
4. Configure the 'Ollama Chat Model' node (named 'Ollama Chat Model') with your Ollama API credentials. It's pre-set for the bespoke-minicheck:latest model.
5. Configure the 'Ollama Model' node (named 'Ollama Model') with your Ollama API credentials. It's pre-set for the qwen2.5:1.5b model. You can change this to another summarization model if preferred.
6. The workflow provides two entry points:
   • 'When clicking ‘Test workflow’': Uses pre-filled sample data in the 'Edit Fields' node for quick testing.
   • 'When Executed by Another Workflow': Allows programmatic execution by passing facts (string: the source document) and text (string: the content to be checked) as JSON input.
7. Review the prompts in 'Basic LLM Chain4' (for fact-checking) and 'Basic LLM Chain' (for summarization) and customize if necessary for specific requirements.
8. Activate the workflow.