Skip to content

v0.1.0

Choose a tag to compare

@vstep1 vstep1 released this 19 May 14:40
· 28 commits to main since this release

First public source release of Local Model Gateway.

Highlights:

  • OpenAI-compatible /v1/chat/completions, /v1/responses, and /v1/models routes.
  • Streamable HTTP MCP endpoint at /mcp with setup and runtime tools.
  • SQLite-backed GPU work queue shared by OpenAI URL requests, MCP runtime jobs, and exclusive llama-cli jobs.
  • Managed local runtime coordination with priority/FIFO admission, runtime swaps, max concurrency, cancellation, and status reporting.
  • Lazy MCP broker package for compact, policy-gated downstream MCP tool discovery.
  • CLI for init, doctor, start, recipes, and service helpers.
  • macOS launchd and generic shell runtime adapter helpers.
  • README demo showing multiple local clients sharing one gateway endpoint.

Validation before release:

  • npm run ci passed locally.
  • GitHub Actions CI passed on main.
  • Public safety scan passed.
  • Workspace package dry-run passed.

This is a GitHub source release. npm publishing is intentionally left as a follow-up.