ggnmem v0.3.1-alpha

Release v0.3.1-alpha: Phase 18 Stretch Goals Complete 🚀

This release finalizes the Phase 18 AI and Knowledge Base capabilities, bringing full support for custom user packs, smarter fallback search, and the BGE-Small model.
ggnmem v0.3.1-alpha expands the offline AI assistant with custom knowledge packs, intelligent fallback search, and full BGE Small ONNX support while remaining fully local, private, and lightweight.

✨ New Features & Enhancements

• Custom Knowledge Packs (mytools.json)
  • Added support for loading custom knowledge packs written as simple, flat JSON arrays (e.g., [{"command": "...", "description": "..."}]).
  • The Knowledge engine now automatically derives the topic name from the file's basename.
• Enhanced Knowledge Diagnostics
  • Rebuilt the ggnmem knowledge list and ggnmem knowledge validate UI to explicitly show loaded file paths for custom packs, entry counts, and specific parsing errors to make debugging custom packs easier.
• Intelligent Search Fallback
  • The ggnmem search Knowledge Base fallback is now much smarter. Instead of only triggering on exactly 0 results, it dynamically evaluates the quality of your history FTS results. If the search only yields low-quality partial matches, Knowledge Base suggestions are seamlessly appended to the bottom.
• BGE Model Downloads (bge-small-en-v1.5)
  • Fully enabled the high-quality BGE Small model! It is no longer restricted as "Coming Soon".
  • You can now seamlessly download the 130MB ONNX model and tokenizer via the ggnmem ai setup interactive CLI wizard.

🛠️ Fixes & Chores

• AI Setup: Fixed a Rust compiler type-mismatch ICE when generating the active embedding provider pipeline.
• Crate Management: Formally wired the new ggnmem-knowledge crate into the root workspace Cargo.toml.
• Formatting: Routine cargo fmt sweeps and cleanup across the CLI and AI crates.