Expand description
Semantic model management (local-only detection).
This module wires the FastEmbed MiniLM embedder into semantic search by:
- validating the local model files
- loading the vector index
- building filter maps from the SQLite database
- detecting model version mismatches
It does not download models. Missing files are surfaced as availability
states so the UI can guide the user. Downloads are handled by [model_download].
Structs§
Enums§
- Semantic
Availability - Unified TUI state machine for semantic search availability.
Functions§
- default_
model_ dir - Get the model directory path for the default MiniLM model.
- default_
model_ manifest - Get the model manifest for the default MiniLM model.
- delete_
vector_ index_ for_ rebuild - Delete the vector index to force a rebuild.
- load_
hash_ semantic_ context - Load hash-based semantic context (no model download required).
- load_
semantic_ context - Load semantic context with optional version mismatch checking.
- load_
semantic_ context_ for_ embedder - load_
semantic_ context_ no_ version_ check - Load semantic context without version checking.
- needs_
index_ rebuild - Check if the vector index needs rebuilding after a model upgrade.