CheckApp is an AI-powered fact-checking system that analyzes claims and generates detailed fact-checking articles with citations from web sources.
- Server Mode: Deploy as a REST API server with job queue management
- Offline Inference: Run fact-checks directly from Python notebooks or scripts
- Evaluation Framework: Benchmark performance of LLMs using the FactCheckingEval dataset
- Multiple Model Support: Works with OpenAI, Google Gemini, OpenRouter, and self-hosted models
- RAG Integration: Retrieval-Augmented Generation for improved source analysis
notebooks/run_workflow.ipynb- Run a complete fact-check in one stepnotebooks/test_stepwiseworkflow.ipynb- Run the workflow step-by-step with customization
| Document | Description |
|---|---|
| Offline Fact-Checking | Guide for running fact-checks locally using Python notebooks |
| Build and Run Docker Server | Deploy the fact-checking API server using Docker |
| Server Specification | API endpoints, request/response formats, and usage examples |
| Fact-Checking Workflow | Detailed explanation of each step in the fact-checking pipeline |
| Supported Models | List of supported AI models and configuration options |
| Running Evaluations | Benchmark LLMs using the evaluation framework |
| Evaluation Results | Performance benchmarks and comparison of different models |
| Model Quantization | Guide for model quantization setup |