Overview

The on-device, offline-first vector-memory engine — a thin Python SDK over a signed native runtime. Install, sign in, and build your first collection here.

Telys is an on-device, offline-first vector-memory engine. The open telys Python SDK drives a closed, signed native runtime. Collections, embeddings, and search run on the machine that imports the SDK — nothing is sent to a server to store or query your data.

The SDK defines the API, verifies release signatures and licenses, and loads the runtime lazily through ctypes. The runtime — vector store, IVF index, and embedding kernels — is provisioned once with telys login, verified offline against embedded public keys, then runs with no network access.

Use Telys for durable semantic memory you control: coding assistants that remember a repository, chat apps that recall past conversations, and RAG pipelines that keep their index and embeddings on the user's hardware. Current version: 0.1.15.

New to Telys? Install the SDK, sign in and activate a license, then walk the Quickstart to create your first collection and run a search.

Start here

Install

Install the SDK with pipx or pip on Python 3.10+.

Sign in

Run telys login and provision the signed runtime.

First collection

Create a collection, add text, and search it end to end.

Verify install

Confirm the runtime is ready and its signature is valid.

Understand Telys

How it works

The thin SDK, the signed runtime, and how they load.

Collections

The core concept — what a collection stores and how.

On-device & offline

Why your data and embeddings never leave the machine.

Build

SDK guides

Create collections, add and upsert, search, filter, tune.

MCP

Give Claude, Cursor, or Codex a memory over 19 tools.

Self-hosting

Run a Telys server and connect with RemoteTelys.

Reference

SDK API index

Every public class and method on the shipped surface.

Install and run your first search

Install the SDK and provision the runtime, then store and query memory locally.

pipx install telys          # or: pip install telys
telys login                 # device-code sign-in; verifies and installs the runtime

Create a collection, add documents, and search — all on-device. The embedder runs in the native kernel that telys login installed.

from telys import Telys
from telys.embedding import AlgentaMultigramEmbedder

db = Telys("./memory")
emb = AlgentaMultigramEmbedder()
col = db.create_collection(
    "notes",
    dim=emb.profile.dimension,        # kernel-sourced multigram dimension
    partition_by="scope",
    embedder=emb,
)

col.add_texts(
    texts=["Telys runs entirely on-device.", "Vectors never leave the machine."],
    ids=["n1", "n2"],
    metadata=[{"scope": "demo"}, {"scope": "demo"}],
)
col.save()                           # commit the collection to disk

hits = col.search_text("where does my data live?", top_k=2)
print(hits["ids"])                   # ['n2', 'n1']

create_collection builds and caches the collection in memory; col.save() is the commit point that writes it to disk. db.collections() lists only saved collections.

Next: Install Telys, then build end to end in the Quickstart.