SQLite extension providing semantic memory with hybrid search.
Field note
SQLite-Memory is a SQLite extension that enables semantic search over text content using vector similarity and full-text search. It chunks text, generates embeddings via local llama.cpp or vectors.space remote service, and supports hybrid search.
Capabilities
Tags
Ways to use it
Python / https://docs.python.org/3/library/sqlite3.html
Not filed / https://docs.sqlitecloud.io/docs/sqlite-memory-examples
Product features
Content-hash based change detection to avoid redundant embedding computation.
Initializes network sync with a managed database.
Pushes and pulls changes for bidirectional sync.
Combines vector similarity with FTS5 full-text search.
Markdown-aware parsing preserves semantic boundaries.
Performs two-phase sync of a directory.
Adds file content, atomically replacing old entries on change.
Adds text content with hash-based deduplication.
Enables CRDT sync for all or a specific context.
Generates embeddings for received content.
Search memory semantically with ranking.
Set API key for remote embedding provider.
Configure embedding model (local or remote).
Chunks text using markdown-aware semantic parsing.
Content-hash change detection skips unchanged files, atomically replaces modified ones, and cleans up deleted ones.
Command-line interface for project initialization, adding sources, searching, and configuration.
Sync operations wrapped in SAVEPOINT transactions for rollback safety.