Release Notes 0.5.30

This page summarizes the product and engineering changes included in 0.5.30, following v0.5.23.

Summary

Release 0.5.30 is the biggest runtime and product consolidation since the last tagged release. It adds a new core automation and continual-learning layer, expands memory and recall workflows, introduces two bundled skills, and hardens the desktop/runtime release path. The broad theme is the same across the codebase: more of CoWork's always-on intelligence is now explicit, persisted, and visible in the UI instead of being scattered across legacy heartbeat, persona, and ad hoc task paths.

New Features

• Core automation profiles: a new automation profile model now owns the always-on runtime surface, replacing older heartbeat/pseudo-persona ownership paths.
• Trace, failure, and learning pipelines: the core runtime now has first-class trace, failure-record, failure-cluster, eval-case, harness-experiment, learnings, memory-candidate, and regression-gate services and repositories.
• Memory distillation: the core memory distiller now operates as a dedicated runtime layer with scoped memory resolution, distill-run persistence, and supporting query paths.
• Research vaults (LLM Wiki): first-class workspace-local research vaults inspired by Andrej Karpathy's LLM Wiki concept, with deterministic raw-source capture, Obsidian-friendly notes, vault search, graph reporting, and filed-back outputs. Learn more
• Programmatic technical video skill (manim-video): a bundled skill for planning and scaffolding Manim CE explainers, equation walkthroughs, algorithm visualizations, and animated architecture/data stories. Learn more
• Curated memory and recall services: new curated-memory, quote-recall, and session-recall services expand how the app stores and surfaces durable context.
• Workspace path healing: desktop workspace paths can now be repaired when a workspace is moved or renamed.
• File provenance registry: imported or exported files can carry trust and provenance metadata through a dedicated registry.
• PDF text extraction: a unified PDF text path now handles normal PDFs and review-heavy extraction flows more reliably.
• Shared onboarding model: onboarding data now lives in shared structures so setup flows can reason about assistant profile, workspace, and response-style state consistently.

Enhancements

• Mission Control and onboarding: the UI now surfaces automation profiles, distillation controls, core failures, learnings, and companion inbox state more clearly.
• Renderer and completion surfaces: task completion UX, disclosure handling, memory hub settings, Slack settings, permission settings, and onboarding screens were updated to fit the new runtime model.
• Gateway and messaging routing: WhatsApp command handling, channel routing, and email/report delivery were updated to match the richer runtime and task-state model.
• Tooling and skill execution: explicit skill invocation matching, runtime tool metadata, tool exposure, and tool-policy inference were refined across the executor and loader layers.
• LLM provider behavior: OpenRouter and provider-factory behavior were tightened with better catalog loading, logging, and model-selection handling.
• Security and approval flows: approval controls, workspace rule handling, control-plane sanitization, and export-permission context handling were all strengthened.
• Documentation: release notes, changelog surfaces, architecture docs, feature docs, security docs, and onboarding guidance were refreshed to reflect the new architecture and release scope.
• Testing: coverage expanded across core automation, memory, onboarding, gateway, runtime tools, security, renderer helpers, and release-adjacent workflows.

Fixes

• Legacy task migration: fixed legacy task-event migration paths so older task data upgrades cleanly into the newer runtime model.
• Heartbeat compatibility: preserved compatibility while routing dispatch state through automation profiles and handling deferred heartbeat state more carefully.
• Completion evidence: fixed report-task contract inference and completion evidence handling so task finalization is more reliable.
• OpenRouter noise: reduced noisy provider catalog loading and made structured logging more predictable.
• Tool-call sanitization: tightened tool-call and task-message sanitization so malicious or malformed payloads are less likely to leak into runtime state.
• Workspace boundaries: edits to workspace ignore files now respect workspace boundaries more strictly.
• Approval and runtime edge cases: updated approval handling, tool exposure, and runtime policy decisions so long-running tasks and guarded actions fail more predictably.
• Release artifacts: release artifact naming and packaging logic were aligned with updater metadata so published assets stay consistent.

Upgrade Notes

• LLM Wiki is the new canonical name for the workspace-local research vault feature; it is inspired by Karpathy's LLM Wiki concept rather than being a general note-taking editor.
• Users upgrading from earlier builds should expect the new automation-profile-driven runtime model to appear in Mission Control and related settings surfaces.
• Existing memory and task data remains compatible, but the release includes a substantial migration and reconciliation layer for older task, heartbeat, and subconscious records.

References

• Changelog - full version history
• Core Automation - current automation runtime model
• LLM Wiki - research vault workflow and graph reporting
• Manim Video - technical explainer video skill
• Memory Flow - current workspace memory architecture

This page is the canonical high-level summary for the changes included in 0.5.30.