OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

Constellation Engine

Constellation Engine is a self-hosted cognitive memory runtime for stateful AI agents. It represents persistent knowledge as typed graph nodes and weighted edges, uses spreading activation to select related memories, compiles them into structured context for an LLM, and supports episodic recall and post-turn consolidation.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Constellation Engine — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when an agent must retain structured knowledge, preferences or interaction history across sessions, and you want that memory to be maintained outside the LLM context window.

Who it is for

  • Developers building long-lived AI agents that need memory across sessions.
  • Agent operators who want a local-first memory layer separate from the model.
  • Teams evaluating an AGPL-3.0 JavaScript runtime for graph-based recall and memory consolidation.

Documented uses

  • Add persistent graph-based long-term memory to an agent.
  • Activate connected memories from a prompt signal and assemble them into context for an LLM.
  • Store and recall episodic interactions across agent sessions.
  • Consolidate agent memories after turns.
  • Keep agent memory separate from the selected LLM.
Before you choose

Requirements & limits

The project is open source under AGPL-3.0. Developer setup requires Node.js 20+ and a configured LLM provider or compatible local endpoint. The README also documents an installer and desktop setup wizard.

  • The runtime is self-hosted and documented as one-owner-per-instance; multi-person serving requires separate instances and an external routing layer.
  • Memory is stored through a knowledge graph and activation process; embedding search is used as an activation signal rather than the sole retrieval stage.
  • SQLite persistence is bundled, and BGE-M3 embeddings run in-process on CPU by default; GPU embedding is configurable.
  • The visual dashboard is part of the official packaged Electron build, while a source-tree run provides a headless engine with minimal dashboard stubs.
  • The AGPL-3.0 license matters when assessing adoption and distribution.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the engine from source

Install and configure the documented headless engine and dashboard for local use.

Before you start

  • Node.js 20+
  • At least one provider key for Anthropic or OpenAI, or a local Ollama/Codex-compatible endpoint
  • A cloned project repository

Steps

  1. Run the command below in the repository.

    Documented command
    npm install

  2. Copy .env.example to .env and configure at least one provider key or compatible local endpoint.

  3. Run the command below to boot the engine and dashboard at http://127.0.0.1:18800.

    Documented command
    npm start

  4. Optionally start the desktop shell with cd electron && npm install && npm start.

Expected outcome

The source-tree engine starts with its runtime and a headless dashboard; the visual dashboard is available only in the official packaged Electron build.

Guide limitations

  • This source setup requires Node.js 20 or later.
  • The source-tree dashboard UI is a minimal stub; it is not the visual dashboard shipped in the packaged Electron build.
  • The documented setup is one owner per instance; multi-person service requires separate instances and an external routing layer.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
CONSTELLATION-ENGINE/constellation-engine · Official homepage title · retained in its source language
How we review sourcesSuggest a correction