Skip to content

v0.11.2

Choose a tag to compare

@github-actions github-actions released this 07 Aug 23:42
· 2964 commits to main since this release
f1f295d

Open Science v0.11.2

Kernel image-output previews for R and Python notebooks, an offline detection indicator with a Network status panel, Task API run-progress liveness and explicit cancellation, isolated unavailable custom MCP discovery, artifact finalization proof-failure diagnostics, and Windows updater certification stabilization — while a broad ownership refactor continues across artifact provenance, session persistence, project files, uploads, and session store.

Open Science is an open-source, model-agnostic AI workbench for scientific discovery — a self-hosted desktop app pairing a plan-and-execute agent with persistent compute, durable project/session storage, and inspectable research artifacts.

v0.11.2 is a patch release on top of v0.11.1. It brings inline image-output previews for R and Python notebook kernel runs — hydrating captured figures and saved plots through a shared output presentation across notebook preview, session dialogs, and expanded conversation tool rows — adds an offline detection indicator in the home header and workspace sidebar with a dedicated Network status panel that reuses the onboarding reachability probe, adds a provider-neutral Task API run-progress heartbeat with a ten-second liveness signal before the first visible output, introduces explicit run cancellation with terminal status and partial-output preservation, isolates unavailable custom MCP server discovery so one broken server does not abort the remaining connector skill synchronization, adds stable internal reason codes to artifact finalization proof rejections for production diagnostics, stabilizes the Windows updater certification handoff through a native process observer and authenticated bootstrap version check, and parallelizes full coverage checks in CI. Under the hood it continues the artifact-provenance ownership refactor — extracting version writer, execution evidence, producer capture, message finalization, persisted execution proof, read model, finalization recovery, staging recovery, unindexed recovery — alongside session-persistence state owner extraction, project-files query and mutation owner extraction, uploads active-transfer and publication owner extraction, session-store message-graph and run-projection owner extraction, and artifact compatibility and publication owner extraction.

✨ Highlights

  • Kernel image-output previews. R and Python notebook runs now hydrate captured figures and saved plots locally in the renderer by compact run ID, sharing one output presentation across notebook preview, the session notebook dialog, and expanded conversation tool rows. R base-graphics, ggplot2, grid, and lattice figures are all rendered. Saved PNG, JPEG, WebP, GIF, AVIF, and SVG images are previewed, TIFF files are decoded, and PDF output renders a first-page preview without cropping. Figure bytes stay in local notebook run state and are never copied into session messages or agent context. (#915)
  • Offline detection indicator and Network status panel. A red offline indicator now appears in the home header and workspace sidebar when the network is unavailable, with a tooltip and click-through to a new Network status panel in Settings. The panel reuses the onboarding reachability probe — a real HTTPS HEAD against known endpoints — to distinguish Connected, Unreachable, and Offline states, shows connection type and local IP, and offers adaptive troubleshooting with a Check again button. The indicator disappears automatically when connectivity returns. (#817)
  • Task API run-progress liveness. A provider-neutral run.progress contract with ordered lifecycle phases and elapsed time emits a ten-second heartbeat until the first visible provider output or terminal run state, so a live but slow provider no longer looks indistinguishable from a stalled request. Progress streams through the public Task WebSocket and is exposed in the SDK, CLI output, and JSONL. (#931)
  • Explicit run cancellation. Headless task runs now have a server-side cancellation operation. The idempotent cancelRun drains provider work, session persistence, and artifact finalization, sets a terminal cancelled status with timestamps, and preserves partial assistant output and successfully finalized artifacts. It is exposed through the HTTP API, SDK, CLI run cancel, and an opt-in --cancel-on-timeout. (#914)

🚀 New Features

  • Kernel image-output previews — full notebook runs hydrate locally in the renderer by compact run ID; one shared output presentation across notebook preview, session dialog, and expanded conversation tool rows; text output independently collapsible while every figure renders in its own always-visible frame; captured figures and unique saved image paths from the run are both presented; PNG/JPEG/WebP/GIF/AVIF/SVG preview, TIFF decode, and PDF first-page render; figure bytes stay in local run state and managed preview resources. (#915)
  • Offline detection indicator and Network status panel — a renderer network store detects connectivity via browser online/offline events; while offline a red indicator appears in the home header pill and workspace sidebar footer with a tooltip and click-through to Settings; the Network panel reuses the onboarding check-row component and shows connection state, connection type (Wi-Fi/Ethernet best-effort), local IPv4 address, and a real reachability probe that distinguishes Connected, Unreachable, and Offline; adaptive troubleshooting suggestions and a Check again button are shown in failed states; recovery re-probes automatically. (#817)
  • Task API run-progress liveness — additive run.progress contract with ordered lifecycle phases and elapsed time; ten-second heartbeat until first visible provider output or terminal run state; provider acceptance observed through the shared ACP prompt lifecycle across Claude Code, OpenCode, Codex Responses, and Codex Bridge; visible output correlated by session and prompt message ID to avoid cross-run contamination; streamed through the public Task WebSocket, SDK, CLI, and JSONL. (#931)
  • Explicit run cancellation — idempotent cancelRun(runId) lifecycle in TaskRunner routed through the existing ACP session owner; waits for provider work, session persistence, and artifact finalization to drain; terminal cancelled status with cancelRequestedAt and cancelledAt timestamps; partial assistant output, tool activity, and finalized artifacts preserved; exposed via HTTP API, SDK cancelRun, CLI run cancel, and opt-in --cancel-on-timeout. (#914)

🔧 Improvements

  • The artifact-provenance ownership refactor continues, extracting version writer, execution evidence, producer capture, message finalization, persisted execution proof, read model, finalization recovery, staging recovery, and unindexed recovery so each lifecycle stage is isolated behind independently testable contracts. (#916, #930, #932, #933, #937, #938, #939, #940, #941, #942)
  • The session-persistence, project-files, uploads, session-store, and artifacts ownership refactor continues, extracting state owner, query and mutation owners, active-transfer and publication and resolver and legacy-recovery owners, message-graph and run-projection and persistence owners, and compatibility and publication owners. (#895, #896, #897, #898, #901, #903, #904, #905, #906, #907, #909, #910, #911, #945, #946)
  • Full coverage checks in CI are now parallelized, nightly and release workflows are optimized, module shadow evidence is published, and README SEO/GEO polish adds answer-first section leads. (#893, #908, #935, #944)

🐛 Bug Fixes

  • Artifact finalization proof-failure diagnostics — every proof rejection now carries a stable internal reason code logged alongside the existing stage and safe ownership identifiers, so production failures can be localized without reproducing the run; error-message redaction is preserved. (#934)
  • Isolated unavailable custom MCP discovery — an enabled custom MCP server whose stdio command target no longer exists no longer aborts the remaining connector skill synchronization; listTools failures are isolated to the unavailable server, stale generated skill documentation is removed for failed servers, and healthy connector skills continue to materialize. (#902)
  • Windows updater certification stabilization — the updater certification now drives the published previous release through its authenticated loopback web RPC without attaching a debugger, attaches a native Windows process observer to the detached NSIS executable, uses the launched application's authenticated bootstrap as the authoritative version check, and redacts token query values from diagnostics. (#899)

📦 Install

Requirements: macOS 12+ (Apple Silicon or Intel), Linux x64, or Windows 10/11 x64. On first run, the onboarding wizard checks the environment and can install and configure an app-managed agent runtime. Once installed, the app can update itself in place.

Download the appropriate package from the Assets section below:

Platform Package
macOS (Apple Silicon) DMG for ARM64
macOS (Intel) DMG for x64
Linux AppImage or Debian package for x64
Windows Installer for x64

macOS — first launch. Official release builds are Developer ID signed and notarized by Apple, so they open like other trusted applications. A locally built copy is not notarized and may require approval through macOS Privacy & Security.

Windows — first launch (unsigned build). No Authenticode certificate yet, so SmartScreen shows a bypassable "unrecognized app" prompt (More info → Run anyway). Verify that the package came from the official release page before continuing.

Build from source instead:

npm install
npm run build:mac   # or: build:linux / build:win

🧭 What's in this release (maturity)

  • ✅ Implemented: a local-first desktop, localhost-web, headless, CLI, and task-SDK surface over persistent projects and sessions with selectable message branches, branching into a new session, and reversible archiving; selectable Claude Code, OpenCode, and Codex agent frameworks behind a shared provider turn-adapter interface; review-gated session plans with durable execution contracts and persistent plan call records; hot-switching of compatible models and providers without reconnecting the agent process; agent-aware context replay respecting each framework's context path; multi-provider model configuration with combined model/reasoning selection; persistent Python/R/REPL kernels and remote SSH execution with read-only installed-package inventories and inline image-output previews; immutable artifact versions with inspectable audit evidence and on-demand LLM-driven code reconstruction; multi-format previews including TIFF and code syntax highlighting; file-based skills with conversational imports, keyword search from GitHub URLs, and ZIP export; permissioned scientific connectors with standard remote OAuth sign-in and portable configuration import/export; opt-in review and bounded correction; personal specialist agent profiles with scoped capabilities, conversational customization, immediate in-flight handoff, and package import/export; scoped permission management with global, project, and session grants and live profile changes during a running turn; conversation export as Markdown and PDF; selective session-artifact downloads; per-turn token usage; message timing metadata; tool-activity elapsed time; completed-turn agent and model identification; context-usage persistence across restarts; a project-scoped command palette; collapsible side panels; split-view file preview beside the session; prompt history navigation; session link favicons; branded loading indicators; adaptive macOS app icons; offline detection indicator with a Network status panel; task API run-progress liveness and explicit cancellation; mobile remote access through Remote.It pairing; configurable storage; diagnostic coverage for user-visible failure boundaries; and installers for macOS, Windows, and Linux.
  • 🚧 Partial: R remains managed-only; provider choice remains constrained by the active framework's endpoint compatibility; remote compute is SSH-only; skills remain local; specialist roles are personal-only (no cross-machine sharing); and review is opt-in and record-scoped.
  • 🗺️ Roadmap: a unified model gateway, a public skills commons and cross-machine specialist sharing, Slurm/cloud-GPU execution, stronger sandboxing and credential isolation, and collaborative research workflows.

🐢 Known Limitations

  • R is managed-only. A bring-your-own R interpreter path is not built yet.
  • Remote compute is SSH-only. Slurm, cloud GPU, and parallel sub-agent fan-out are not built yet.
  • Provider choice is per framework, not one unified gateway. The available protocol depends on the selected agent backend.
  • Hot-switching applies only to registered compatible targets. Framework, auth-lane, wire-route, or unsafe capability changes still require a reconnect.
  • Code reconstruction is LLM-generated. It does not replace deterministic reproduction; portable environment locks and full-fidelity session replay remain open.
  • Specialist roles are personal-only. There is no shared public commons or cross-machine specialist forking yet.
  • The task SDK is a first-generation surface. Task creation, polling, artifact retrieval, run progress, and cancellation work; broader orchestration remains open.
  • Switching agent backends cannot transfer in-flight tool state. Existing conversation history can replay, but a running action is not migrated.
  • Skills are local only. There is no shared public commons, cross-machine forking, or user-facing version pinning yet.
  • The reviewer is opt-in and record-scoped. It does not replace domain-specific validation of citations, units, statistics, or methods.
  • Scoped permissions cover allow-grants only. Network sandboxing, directory-level file access control, and a credential vault are not built yet.
  • Windows builds are unsigned. SmartScreen may warn on first launch; official macOS builds are notarized.
  • No local GPU compute backend.
  • No multi-user real-time collaboration.

🙏 Acknowledgements

Thanks to @ewen-poch, @wen2zhou, @daanveer-tech, @justemu, and everyone in Discord, X, and Discussions.


What's Changed

  • feat(network): offline detection indicator and Network status panel by @ewen-poch in #817
  • fix(macOS-release): stabilize signing and native notarization smoke by @ewen-poch in #891
  • ci(test-impact): publish module shadow evidence by @ewen-poch in #893
  • test(artifacts): characterize storage compatibility by @ewen-poch in #895
  • test(session-store): characterize public contract by @ewen-poch in #896
  • refactor(session-store): extract persistence owner by @ewen-poch in #897
  • refactor(artifacts): extract publication owner by @ewen-poch in #898
  • fix(windows-update): stabilize updater certification handoff by @ewen-poch in #899
  • refactor(artifacts): extract compatibility owner by @ewen-poch in #901
  • fix(connectors): isolate unavailable custom MCP discovery by @ewen-poch in #902
  • test(uploads): characterize repository contracts by @ewen-poch in #904
  • test(project-files): characterize repository contracts by @ewen-poch in #905
  • refactor(session-store): extract message graph owner by @ewen-poch in #903
  • refactor(uploads): extract active transfer owner by @ewen-poch in #907
  • README SEO/GEO polish: answer-first leads by @daanveer-tech in #908
  • refactor(project-files): extract mutation owner by @ewen-poch in #909
  • refactor(uploads): extract publication and resolver owners by @ewen-poch in #910
  • refactor(project-files): extract query owner by @ewen-poch in #911
  • refactor(uploads): extract legacy recovery owners by @ewen-poch in #912
  • fix(task-api): add explicit run cancellation by @ewen-poch in #914
  • feat(notebook): preview kernel image outputs by @ewen-poch in #915
  • test(artifact-provenance): characterize write contracts by @ewen-poch in #913
  • test(artifact-provenance): characterize lifecycle contracts by @ewen-poch in #916
  • refactor(artifact-provenance): extract version writer by @ewen-poch in #930
  • feat(task-api): expose run progress liveness by @ewen-poch in #931
  • refactor(artifact-provenance): extract execution evidence by @ewen-poch in #932
  • refactor(artifact-provenance): extract producer capture by @ewen-poch in #933
  • fix(artifacts): classify finalization proof failures by @ewen-poch in #934
  • ci(pr-gate): parallelize full coverage checks by @ewen-poch in #935
  • refactor(artifact-provenance): extract persisted execution proof by @ewen-poch in #937
  • refactor(artifact-provenance): extract message finalization by @ewen-poch in #938
  • refactor(artifact-provenance): extract staging recovery by @ewen-poch in #939
  • refactor(artifact-provenance): extract finalization recovery by @ewen-poch in #940
  • refactor(artifact-provenance): extract unindexed recovery by @ewen-poch in #941
  • refactor(artifact-provenance): extract read model by @ewen-poch in #942
  • test(artifact-provenance): certify repository architecture by @ewen-poch in #943
  • ci(pipelines): optimize nightly and release workflows by @ewen-poch in #944
  • test(session-persistence): characterize coordinator contracts by @ewen-poch in #945
  • refactor(session-persistence): extract state owner by @ewen-poch in #946

Full Changelog: v0.11.1...v0.11.2