With FTS5 landing in #690, the full-text index can enrich LLM chat context beyond just documents.
Extend FTS to cover all text-heavy entity fields (project titles/notes, vendor notes, maintenance item names/notes, incident titles/notes, etc.) so the chat can surface relevant structured and unstructured data before generating SQL.
When a user asks a question, FTS pre-filters relevant content and injects it into the LLM prompt, giving the model richer context to synthesize answers from.
With FTS5 landing in #690, the full-text index can enrich LLM chat context beyond just documents.
Extend FTS to cover all text-heavy entity fields (project titles/notes, vendor notes, maintenance item names/notes, incident titles/notes, etc.) so the chat can surface relevant structured and unstructured data before generating SQL.
When a user asks a question, FTS pre-filters relevant content and injects it into the LLM prompt, giving the model richer context to synthesize answers from.