Documentation
¶
Overview ¶
Package vec implements a SQLite virtual table for vector search with MATCH semantics. Each virtual table has a per-table shadow table that stores dataset identifiers, document ids, content, metadata, and embeddings. A generic index blob is persisted in the shared vector_storage table, and an in-memory cache accelerates queries.
Features:
- WHERE doc_id MATCH ? using an encoded embedding BLOB
- Optional match_score column for inspecting similarity scores
- Auto-created shadow tables and triggers
- Index persistence in vector_storage and cache invalidation on writes
- Pluggable index (brute-force baseline)
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func InvalidateCache ¶
InvalidateCache clears cached indices for a given shadow/dataset across active connections.
Types ¶
type Cursor ¶
type Cursor struct {
// contains filtered or unexported fields
}
Cursor scans results from a vec table.
type Module ¶
type Module struct {
// contains filtered or unexported fields
}
Module implements vtab.Module for the vec virtual table. It creates a per-table shadow store and supports MATCH-based similarity scans.
type Table ¶
type Table struct {
// contains filtered or unexported fields
}
Table represents a single vec virtual table instance.
func (*Table) Disconnect ¶
Disconnect cleans up per-connection resources.
Source Files