v0.13.1

Open Science v0.13.1

Local folder access via @path mentions, bounded REPL inference, project-scoped frame reads, artifact lineage access, connector runtime status and recovery, refined session sidebar rows with project actions and agent context, and notebook artifact read access — alongside hardened database migrations, Windows upload repair, connector login recovery, and workspace interaction fixes.

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.

v0.13.1 is a patch release. It lets researchers grant local folder access through @path mentions so the agent can read an entire directory without attaching individual files; adds bounded REPL inference so the agent can evaluate short JavaScript expressions in a controlled runtime; introduces project-scoped frame reads giving the agent live access to the active project context; surfaces artifact lineage so version ancestry is traceable from any generated file; brings connector runtime status and recovery into view so sign-in and connection issues are visible and recoverable; refines the session sidebar with project actions and agent context rows; and adds notebook artifact read access so the agent can inspect generated files directly from the notebook runtime. On the fix side it hardens database migrations with a versioned startup gate, legacy check-constraint compatibility, and pre-migration recovery; repairs Windows atomic upload publication; preserves connector-managed login recovery and exposes custom skills across agents; recovers unavailable persisted Codex sessions; makes review suspension durable across transport loss; fails closed on settings read errors; coordinates compute jobs with owner deletion; and resolves workspace interaction priority, question advancement, and choice persistence issues.

✨ Highlights

• Local folder access with @path mentions. Researchers can now type @path/to/folder in the composer to grant the agent read access to an entire local directory, without attaching each file individually. The granted folder is scoped to the session and the agent can browse its contents for the duration of the task. (#798)
• Connector runtime status and recovery. Connector sign-in state, connection health, and recovery flows are now surfaced in the workspace so a broken or expired connector session is immediately visible and can be re-established without restarting the app. (#1085)
• Bounded REPL inference. The agent can now evaluate short JavaScript expressions in a bounded runtime to test hypotheses, transform data, or verify logic inline — without spinning up a full notebook session. (#1097)
• Artifact lineage access. The agent can trace the ancestry of any generated artifact version, walking back through the chain of versions that produced it, so reproducibility questions and provenance audits can be answered from within a session. (#1086)

🚀 New Features

• Local folder access via @path mentions — typing @path/to/folder in the composer grants the agent read access to the referenced directory for the active session, so directory-wide file discovery and reads are possible without individual file attachment. (#798)
• Bounded REPL inference — a bounded JavaScript REPL is available to the agent for evaluating short expressions, testing transformations, or verifying logic in a controlled runtime without a full notebook session. (#1097)
• Project-scoped frame reads — the agent can now read the live project context (frame) to understand the current workspace state, open sessions, and active configuration during a turn, enabling more contextually aware responses. (#1090)
• Artifact lineage access — the agent can access and traverse the lineage of any generated artifact, walking its version ancestry to understand how a result was derived and what preceded it. (#1086)
• Connector runtime status and recovery — connector connection state, sign-in health, and recovery surfaces are now visible in the workspace, with recovery flows for expired or broken sessions. (#1085)
• Refined session sidebar rows — session sidebar rows have been refined with cleaner layout, status indicators, and improved readability for active, pending, and completed sessions. (#1081)
• Project actions and agent context — the session sidebar now surfaces project-level actions and agent context information, giving quicker access to common operations and making the active agent's configuration visible at a glance. (#1067)
• Notebook artifact read access — the agent can now read artifacts generated by the notebook runtime directly, without requiring a separate file read, closing the loop between code execution and artifact inspection. (#1076)

🔧 Improvements

• Database schema migrations are now enforced as versioned in CI, ensuring every migration step is tracked and no ad-hoc schema changes slip through. (#1095)
• Cross-platform CI runner contention has been reduced, improving PR-gate throughput and stability. (#1078)
• The project description has been broadened to reflect general research support across disciplines beyond the life sciences. (#1093)
• The star-history image has been removed from the README to keep the focus on the product. (#1096)

🐛 Bug Fixes

• Connector-managed login recovery preserved — connector sessions using connector-managed OAuth could lose their recovery state on restart, leaving the user unable to reconnect without removing and re-adding the connector. The recovery flow now preserves the managed login state correctly. (#1100)
• Custom skills exposed across agents — custom (personal) skills were not being exposed to all agent frameworks, so a skill created or imported under one backend could be invisible under another. Skills are now consistently available across agents. (#1091)
• Versioned database migration startup gate — the app now validates that all required database migrations have been applied before startup, blocking launch with a clear message if a migration is missing rather than failing later with an opaque error. (#1077)
• Legacy check-constraint compatibility — databases created under an earlier schema with equivalent but differently named check constraints were being rejected by migration validation. The validator now accepts equivalent legacy constraints. (#1089)
• Pre-migration recovery preserved — recovery data captured before a migration could be lost when the migration ran, leaving sessions without their pre-migration state. The migration now preserves recovery records. (#1092)
• Windows atomic upload publication repaired — the atomic publication path on Windows could fail under certain filesystem locking conditions, leaving an upload in a partially published state. The publication path is now repaired and robust against these conditions. (#1075)
• Unavailable persisted Codex sessions recovered — Codex sessions persisted in a state that the runtime could not restore were silently dropped on load. These sessions are now recovered with their conversation history intact. (#1103)
• Review suspension durable across transport loss — a session-plan review suspension could be lost if the transport connection between the agent and the app dropped, causing the plan to resume without the intended review pause. The suspension is now persisted durably. (#1098)
• Provenance and pending review state preserved — session-plan provenance and the pending review flag could be reset on certain lifecycle transitions, losing track of which steps had been reviewed. Both are now preserved across transitions. (#1074)
• Settings fail closed on read errors — a read error in the settings backend could cause the app to start with default settings instead of the user's configured values, silently dropping configuration. The app now fails closed with a visible error instead. (#1080)
• Compute jobs coordinated with owner deletion — compute jobs owned by a deleted session or project could continue running as orphans, consuming resources. Jobs are now coordinated with their owner's lifecycle and cleaned up on deletion. (#1043)
• Workspace interaction priority preserved — pending interactions (ask-user, permission requests) could lose their priority ordering when multiple were queued, causing the wrong interaction to surface first. Priority is now preserved. (#1079)
• Questions require next to advance — structured clarification questions could advance automatically without an explicit next action, skipping questions the user had not yet answered. An explicit next is now required. (#1082)
• Choices persisted only after next — choices in a multi-question clarification card were being persisted immediately on selection rather than after advancing, so a user who changed their mind and navigated away could leave stale choices. Choices are now persisted only after the next action. (#1084)
• Branded database loading animation reused — the startup database loading step was using a generic spinner instead of the branded loading animation used elsewhere. The branded animation is now reused for consistency. (#1094)
• PR gate baseline restored — a CI policy test could produce inconsistent summary output, causing the PR gate to fail spuriously. The summary output is now isolated and deterministic. (#1102, #1099)

📦 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, reversible archiving, project pinning, and persistent side conversations; 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; a unified plan-approval, permission-approval, and ask-user composer lane; 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, inline image-output previews, host capability introspection, notebook artifact read access, and bounded REPL inference; immutable artifact versions with inspectable audit evidence, artifact lineage access, and on-demand LLM-driven code reconstruction; multi-format previews including TIFF and code syntax highlighting; file-based skills with conversational creation, conversational imports, keyword search from GitHub URLs, authenticated GitHub access, and ZIP export; permissioned scientific connectors with standard remote OAuth sign-in, portable configuration import/export, and runtime status and recovery surfaces; 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; local folder access via @path mentions; 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; branch-aware context-window run trends; a project-scoped command palette; refined session sidebar rows with project actions and agent context; project-scoped frame reads; 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; desktop notifications plus durable unread conversation badges; a cross-surface notification message center with durable read state; structured agent clarification cards; live session status on the Home dashboard; and session keyboard shortcuts.
• 🚧 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.