cogxai-local
v2.8.2
Published
Local-first providers for CogxAI — SQLite storage + GTE-Small embeddings, fully offline
Downloads
366
Maintainers
Readme
cogxai-local
Local-first providers for CogxAI — SQLite storage + GTE-Small embeddings. Fully offline, zero API keys.
Install
npm install cogxai-localUsage
import { SqliteProvider, GteSmallEmbeddings } from 'cogxai-local';
const storage = new SqliteProvider('./my-agent-memory.db');
const embeddings = new GteSmallEmbeddings(); // ~30MB download on first run
// Use with cogxai-brain
import { CogxAIEngine } from 'cogxai-brain';
const engine = new CogxAIEngine({ storage, embeddings });
await engine.store({ content: "User prefers dark mode", type: "semantic" });
const memories = await engine.recall({ query: "theme preferences" });Why local?
- Zero API keys — no Supabase, no Voyage, no nothing
- Fully offline — works on a plane, in a bunker, wherever
- Fast — no network latency, sub-millisecond reads
- Private — your data never leaves your machine
- Free — no usage costs, no rate limits
Providers
SqliteProvider
SQLite-backed storage with in-memory vector index. Implements StorageProvider from cogxai-brain.
GteSmallEmbeddings
Local ONNX runtime using Xenova/gte-small (384 dimensions). Downloads ~30MB model on first run, fully offline after.
Part of the CogxAI ecosystem
cogxai-brain— Core memory enginecogxai-cloud— Supabase + Voyage providerscogxai-local— This package. SQLite + local embeddings.cogxai— Full package with MCP server, CLI, SDK, and more.
License
MIT — Black-dork
