Memory 🧠

EngliVec — Engli's pure-TypeScript vector database, running live in this tab. Actions, informations, and objects are stored as vectors; relations are vector edges (holographic bindings), so the graph can be queried with geometry: search, recall, analogy, aggregation.

Loading the knowledge snapshot…

Semantic search

try:filter:

Teach Engli a memory

Stored as a vector in your database (localStorage snapshot), searchable above, optionally linked with a vector edge.

Ask the graph

subject --relation→ ? — explicit edges answer first, then the holographic trace (unbind + nearest-neighbor) answers geometrically.

try: Abandon --risks→ ? · Zip a folder --uses→ ? · research a topic and write a report --deploys→ ?

Aggregate knowledge

Cluster what Engli knows around a topic and pick a diverse, relevant subset (maximal marginal relevance) — context-window-sized knowledge for an agent.

The same engine, as an API

Every query on this page also works over HTTP — and /api/vector/memory is a stateless protocol where an agent owns its memory as a snapshot token: send it, mutate it, get it back. Serverless-native; nothing is stored on the server.

GET  /api/vector/search?q=scrape+website+data&k=5
GET  /api/vector/recall?subject=Abandon&rel=risks
GET  /api/vector/aggregate?q=virtues+of+communication
POST /api/vector/memory
     { "ops": [ {"op":"remember","kind":"action","text":"deployed v2"},
                {"op":"link","from":"deployed v2","rel":"affects","to":"checkout"},
                {"op":"recall","subject":"deployed v2","rel":"affects"} ] }
  -> { "results": [...], "snapshot": "<base64 — your memory, hand it back next time>" }