@loreai/opencode

Experimental — Under active development. APIs, storage format, and behavior may change.

Three-tier memory architecture for OpenCode — distillation, not summarization.

An implementation of Sanity's Nuum memory architecture and Mastra's Observational Memory system as an OpenCode plugin. Preserves operational intelligence (file paths, error messages, exact decisions) rather than narrative summaries that lose the details agents need to keep working.

Install

Add to your project's opencode.json:

{
  "plugin": [
    "@loreai/opencode"
  ]
}

Restart OpenCode and the plugin will be installed automatically.

This package is also published as opencode-lore (legacy alias). Both names ship identical code at every release — either works.

Companion packages

Lore ships as three packages sharing the same SQLite database at ~/.local/share/opencode-lore/lore.db:

• @loreai/opencode (you are here) — OpenCode plugin
• @loreai/pi — Pi coding-agent extension
• @loreai/core — shared memory engine

Switching between OpenCode and Pi on the same project preserves the curated knowledge, distillations, and AGENTS.md sync.

Documentation

Full architecture, benchmarks, configuration, and rationale: github.com/BYK/loreai

License

MIT — see LICENSE.