EquipDoc-Agent v0.1.0

Highlights

• Reproducible model-free Gradio demo
• LangGraph human-in-the-loop tool approval
• Safe .npy signal validation and sandboxing
• Explicit Demo/Full mode boundaries
• RAG lexical fallback when the vector database is unavailable
• Evidence-aware maintenance report generation
• Automated CI on Python 3.10, 3.11 and 3.12

Verification

• 13 unit tests passed
• Strict health check passed
• Demo smoke test passed
• GitHub Actions passed on all supported Python versions

Scope

This release demonstrates the complete Agent workflow without requiring model weights or a GPU. The reported fault type in Demo mode is a clearly labelled fixed fixture and must not be interpreted as a real diagnostic result.

Full-model evaluation, cross-condition Group Split, groundedness review and deployment benchmarks are planned for P1.