Skip to content

v0.12.0

Choose a tag to compare

@github-actions github-actions released this 08 Aug 16:23
· 2935 commits to main since this release
c0fcbe3

Open Science v0.12.0

A cross-surface notification message center, a structured agent clarification workflow, authenticated GitHub skill imports, live session status on the Home dashboard, session-resume continuation for interrupted turns, and notebook output-artifact separation — while a broad workspace-runtime and ownership refactor continues across session persistence, project files, inputs, and transport contracts.

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.

v0.12.0 is a minor release. It introduces a backend-owned notification message center that unifies task outcomes and authorization requests across desktop, web, and mobile with durable read state; adds a structured clarification workflow so agents can ask multi-question requests through app-owned blocking choice cards instead of falling back to raw Markdown or permission prompts; brings authenticated GitHub skill imports to lift the unauthenticated rate-limit ceiling on repository search, scan, and preview; surfaces live session status — running, waiting-on-you, and newly completed — on the Home dashboard with horizontally scrollable update cards; continues an interrupted agent turn after Resume without duplicating the original user message; and separates notebook kernel-captured output from saved artifacts so historical cells render their original bytes instead of later file mutations. Under the hood it continues the workspace-runtime ownership refactor — extracting runtime event, prompt preparation, message command, and session lifecycle owners — alongside project-files-view query, preview, and presentation owner extraction, session-persistence reconciliation and deletion owner extraction, unified immutable input authority, and deepened application-renderer transport contracts.

✨ Highlights

  • Cross-surface notification message center. A backend-owned, bounded notification inbox projection now unifies task outcomes and blocking authorization requests across desktop, web, and mobile surfaces with durable read state, authorization action state, retry-safe deduplication, and structured project/session navigation targets. One shared notification store and bell replaces the legacy desktop-only unread-task runtime, and native desktop banners and badges continue as optional delivery adapters. (#950)
  • Structured agent clarification workflow. Agents can now request structured multi-question clarifications through app-owned blocking choice cards with previous/next navigation, selection states, custom answers, Skip, and Let the agent decide. Pending questions are persisted as session activities with owning-runtime metadata, the active agent turn ends while waiting, and the owning framework and provider resume after the answer with history replay when provider context was reset. The workflow is preserved across Claude Code, OpenCode, Codex Responses, and Codex Bridge adapters. (#976)
  • Authenticated GitHub skill imports. A compact GitHub token control is now available in the Import from GitHub header with verify/save, replace, masked status, failure, and remove states. Verified tokens are persisted as safeStorage-encrypted references and applied as Bearer authorization only to exact api.github.com and raw.githubusercontent.com hosts, lifting the unauthenticated rate-limit ceiling for repository search, direct scan, lazy preview, Settings import, and conversation-requested import. (#963)
  • Live session status on Home. The Home dashboard now surfaces live running and waiting-on-you counts per project, horizontally scrollable session-update cards for running, needs-you, and unread completed sessions, project names as the source for update cards and recent sessions, aligned status colors and motion, and a global search button routed through the existing command palette flow. Completed cards are removed after their result has been read, and the layout is preserved across Electron, web, and compact mobile widths. (#973)

🚀 New Features

  • Cross-surface notification message center — a backend-owned, bounded NotificationInboxItem projection with durable read state, authorization action state, retry-safe deduplication, and structured project/session navigation targets; user-initiated task outcomes and blocking authorization requests are projected into the inbox while ordinary project/session management events are excluded; replaces the legacy unread-task runtime with application commands, Electron IPC adapters, web renderer contracts, and startup migration; one shared notification store and bell on Home, desktop Workspace, the mobile conversation header, and the Workspace sidebar; the mobile message center is presented as a bottom drawer with automatic sidebar collapse; native desktop banners and badges remain as optional delivery adapters. (#950)
  • Structured agent clarification workflow — application-owned structured elicitation contract with detached continuation workflow; pending questions, draft answers, completed answers, and owning-runtime metadata persisted in the existing session JSON activity model; the active agent turn ends while waiting, then resumes the owning framework and provider after the answer with history replay when provider context was reset; multi-question requests presented sequentially in one blocking bottom card with previous/next navigation, selection states, custom answers, Skip, and Let the agent decide; a pending placeholder and completed read-only answer summary are kept in the transcript without rendering the underlying call as permission or failed tool activity; pending questions surface through the notification flow; behavior preserved across Claude Code, OpenCode, Codex Responses, and Codex Bridge; responsive sizing, draggable card height, narrow-screen behavior, hover/focus affordances, and keyboard navigation. (#976)
  • Authenticated GitHub skill imports — compact GitHub token control in the Import from GitHub header with verify/save, replace, masked status, failure, and remove states; candidate tokens verified with GitHub before replacing an existing credential; persisted as a safeStorage-encrypted reference with masked display; Authorization Bearer applied only to exact api.github.com and raw.githubusercontent.com hosts; authenticated fetch seam reused for repository search, direct scan, lazy preview, Settings import, and conversation-requested import; GitHub rate-limit failures point users back to the token control; anonymous behavior preserved when no decryptable token is configured. (#963)
  • Live session status on Home — Home search button routed through the existing command palette flow; live running and waiting-on-you counts derived per project and refreshed while Home remains mounted; horizontally scrollable session-update cards for running, needs-you, and unread completed sessions; completed cards removed after the result has been read; project names shown as the source for update cards and recent sessions; aligned status colors and motion — blue animated running, amber needs-you, green completed; refined Projects icon, card borders, scroll gutters, project action menu, Project Settings copy, notification completion icon, and theme menu wrapping; layout preserved across Electron, web, and compact mobile widths including 320, 375, 390, 414, and 768 px. (#973)

🔧 Improvements

  • The workspace-runtime ownership refactor continues, extracting runtime event owner, prompt preparation owner, message command owner, and session lifecycle owner so each runtime responsibility is isolated behind independently testable contracts. (#962, #969, #971, #972)
  • The project-files-view ownership refactor continues, extracting preview owner, query model owner, and presentation owner. (#961, #964, #965)
  • The session-persistence ownership refactor continues, extracting reconciliation owner and deletion owner. (#948, #951)
  • Unified immutable input authority and deepened application-renderer transport contracts isolate input lifecycle and inter-process communication behind stable boundaries. (#966, #967)
  • Expanded visual-regression and accessibility E2E coverage and normalized session-store architecture diagnostic paths. (#957, #958, #970, #974, #975)
  • CONTRIBUTING guidance now explicitly allows module-scoped local tests. (#955)

🐛 Bug Fixes

  • Session-resume continuation without duplicate messages — Resume now continues an interrupted agent turn without removing and re-appending the unanswered user prompt, eliminating duplicate user messages and repeated work. Exactly one original user message is kept with an interrupted marker, a hidden app-owned continuation is reconstructed from durable session state using suppressUserMessage so it cannot publish another visible user event, and the durable original request and prompt media are included so a disconnect before provider acceptance cannot lose the task. If the provider session resumes natively the hidden continuation enters the restored context; if native resume is unavailable the session fresh-adopts with a bounded active-branch transcript replay before the continuation. (#936)
  • Live permission-wait preservation — an additional sessions reload no longer normalizes a waiting-permission state as an app-restart interruption while the owning ACP prompt is still alive. The session repository gains a late-bound view of active prompt ownership and preserves runtime status, messages, activities, activity groups, and conversation-graph state only while the exact owning prompt remains active; genuine restart normalization still applies when no live prompt owns the session. (#968)
  • Notebook output-artifact separation — kernel-captured run outputs are now the sole authority for notebook and message cell figures; saved-only files are no longer resolved as notebook output, preventing historical cells from showing later file mutations after an app restart. Jupyter export draws cell outputs only from run outputs, keeping artifacts as independent files. R plots drawn to built-in file devices are captured by replaying R's native display list into the kernel PNG device, covering base graphics, ggplot2, grid, and lattice file-device output. (#952)

📦 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, authenticated GitHub access, 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; a cross-surface notification message center with durable read state; structured agent clarification workflow; live session status on the Home dashboard; session-resume continuation for interrupted turns; and notebook output-artifact separation
  • 🚧 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

  • fix(session-resume): continue interrupted turns without duplicate messages by @ewen-poch in #936
  • ci(windows): include updater metadata in upgrade smoke by @ewen-poch in #949
  • feat(notifications): add cross-surface message center by @ewen-poch in #950
  • refactor(session-persistence): extract reconciliation owner by @ewen-poch in #951
  • fix(notebook): separate captured output from saved artifacts by @ewen-poch in #952
  • test(session-persistence): certify coordinator ownership by @ewen-poch in #953
  • test(session-store): certify ownership boundaries by @ewen-poch in #954
  • docs(contributing): allow module-scoped local tests by @ewen-poch in #955
  • test(workspace-runtime): characterize hook ownership boundaries by @ewen-poch in #956
  • test(project-files-view): characterize interaction boundaries by @ewen-poch in #957
  • test(session-store): normalize architecture diagnostic paths by @ewen-poch in #958
  • refactor(project-files-view): extract preview owner by @ewen-poch in #961
  • refactor(workspace-runtime): extract runtime event owner by @ewen-poch in #962
  • feat(skills): add authenticated GitHub imports by @ewen-poch in #963
  • refactor(project-files-view): extract query model owner by @ewen-poch in #964
  • refactor(project-files-view): extract presentation owner by @ewen-poch in #965
  • refactor(inputs): unify immutable input authority by @ewen-poch in #966
  • refactor(transport): deepen application renderer contracts by @ewen-poch in #967
  • fix(session-persistence): preserve live permission waits by @ewen-poch in #968
  • refactor(workspace-runtime): extract prompt preparation owner by @ewen-poch in #969
  • test(e2e): expand visual and accessibility states by @ewen-poch in #970
  • refactor(workspace-runtime): extract message command owner by @ewen-poch in #971
  • refactor(workspace-runtime): extract session lifecycle owner by @ewen-poch in #972
  • feat(home): surface live session status by @ewen-poch in #973
  • test(workspace-runtime): certify facade composition by @ewen-poch in #974
  • test(workspace-runtime): reject unresolved owner loaders by @ewen-poch in #975
  • feat(agent-input): add structured clarification workflow by @ewen-poch in #976
  • refactor(session-persistence): extract deletion owner by @ewen-poch in #948

Full Changelog: v0.11.2...v0.12.0