### Key Updates:
* Enhanced Hallucination Detection* Dingo 1.9.0 integrates two powerful hallucination detection approaches: - *HHEM-2.1-Open local model* (recommended) - runs locally without API costs - *GPT-based cloud detection* - leverages OpenAI models for detailed analysis
Both evaluate LLM-generated answers against provided context using consistency scoring (0.0-1.0 range, configurable thresholds).
* Configuration System Overhaul* Complete rebuild with modern DevOps practices: - Hierarchical inheritance (project → user → system levels) - Hot-reload capabilities for instant config changes - Schema validation with clear error messages - Template system for common scenarios
* DeepWiki Document Q&A* Transform static documentation into interactive knowledge bases: - Multi-language support (EN/CN/JP) - Context-aware multi-turn conversations - Visual document structure parsing - Semantic navigation and cross-references
### Why It Matters: Traditional hallucination detection relies on static rules. Our approach provides context-aware validation essential for production RAG systems, SFT data quality assessment, and real-time LLM output verification.
Perfect for: - RAG system quality monitoring - Training data preprocessing - Enterprise knowledge management - Multi-modal data evaluation
*GitHub*: https://github.com/MigoXLab/dingo *Docs*: https://deepwiki.com/MigoXLab/dingo
What hallucination detection approaches are you currently using? Interested in your RAG quality challenges.