Meeting 12: Visualizing AI Truth in OllamaModelTester (Technical)

What started as a Monday mission is now a deployed, tested and functional reality. The OllamaModelTester has been transformed into a transparent, configurable and visually interpretable AI evaluation platform. The AI-generated sentence is now evaluated against the source using Natural Language Inference (NLI) and color-coded in real time. You can now choose which NLI model completes the truth analysis. The UI now displays INTEGER and FLOAT metrics, focusing on what matters on AI model and human evaluation comparison. By making truth visible, measurable and configurable, OllamaModelTester now tests AI models and holds them accountable. Chapters: 0:00 - Presentation 6:32 - Add Sentence-Level Truth Analysis in OllamaModelTester 28:34 - Build and Test Sentence-Level Truth Analysis 36:07 - Create a GitHub Pull Request 43:31 - Create a GitHub Release 46:11 - Upload Package and Create a New PIP Release 47:49 - Add sentence scores import and export module 50:35 - Create a New Pull Request 57:16 - Create a New GitHub Release 29:25 - Upload a New Package and Create PIP Release 1:00:41 - Modify FastAPI Backend 1:10:40 - Modify React Frontend 1:43:52 - Test and Confirm Web Platform Successful Workflow 2:03:58 - Create OllamaModelTester Web Pull Request 2:12:58 - OllamaModelTester LIVE Deployment and Test Actions completed: • Add Sentence-Level Truth Analysis – Done • Configure Default Fingerprint Metrics – Done • Add Dynamic NLI Cross-Encoder Model Selection – Done • Add sentence scores import and export module – Done • Update and publish a new Python package – Done • Deploy the OllamaModelTester LIVE web – Done • Create a technical presentation – Done • Finalize class – Done • Upload files in GitHub repository – Done