plompy-engine.Docs
Historical releaseAll 16 releases ↗

Plompy 0.4.0

Provider-bound runners and offline demos

This is a historical release. Its notes describe this version, not the current setup. See the latest stable release before starting a new installation.

This version predates Plompy Local. It covers the Discord engine; the separate local MCP gateway and starter kit arrive in 0.13.0.

GitHub notes and assets for this version may use the earlier Szpont name. Historical filenames and source links are preserved.

What changed#

Runner adapters#

  • Add experimental Claude Code and OpenCode host runners alongside the default Codex CLI, with explicit deployment-level selection for new threads.
  • Keep provider readiness and queue eligibility independent, without automatic engine or model fallback when another runner is unavailable.
  • Validate exact provider sessions, explicit completion events, bounded subprocesses and typed work checkpoints. Preserve cancellation reasons and process-group cleanup.
  • Isolate provider configuration variables and private authentication directories. Add runner-aware diagnostics and public-tree checks for private authentication/session paths.
  • Document provider-specific permissions, login, model and MCP configuration. OpenCode requires explicit acknowledgement of host execution; its local credential check does not prove remote authentication.

Durable state and upgrade boundary#

  • Advance public SQLite state to schema 3 with immutable conversation bindings and job provider snapshots. Legacy conversations, sessions and queued work remain Codex independently of the configured default.
  • Preserve provider bindings across restart, manual recovery and session reset. A default-runner change applies only to new threads; uncertain work is never rerouted or replayed automatically.
  • Extend verified state backup and restore support to schema 3. Rollback to 0.3.x requires a matching pre-upgrade snapshot and reconciliation of later actions.
  • Verify migrations, rollback, scheduling and adapters with offline synthetic regressions. No live Claude/OpenCode account, model or host-sandbox acceptance is claimed by these tests.

Documentation and offline demo#

  • Rework the README around concrete use cases, runner support levels and deployment boundaries, with a reproducible three-scenario gallery.
  • Add an account-free demo using the actual Engine and SQLite Store with scripted transport and replies. Include raw transcripts and separate optional capture tooling.

Distribution boundary#

  • Refresh the README banner and social-preview asset to reflect Codex, Claude Code and OpenCode while retaining the project identity.
  • Keep the published Docker image Codex-only. Reject alternative runner selection before initialization writes until independent container acceptance exists.
  • Preserve the actual Codex sandbox and the non-root, read-only image acceptance path on Linux amd64 and arm64.

Downloads and source#

This GitHub release has no separately uploaded download assets. Its source and published distribution details are linked from the release page.

Original GitHub release · Source at v0.4.0 · Changelog at this tag

Before you upgrade#

Back up private state and credentials, review the version's migration boundaries, and retain a compatible rollback snapshot. A release does not update an existing deployment automatically.

Full release history · Upgrade and rollback guide

Find your next step

Search titles and guide content.