AIPOCH Open-Science v0.27.0

AIPOCH Open-Science v0.27.0

A release that scales the literature workspace and moves long work to the background: batch PDF import with a broad literature stability pass, durable background execution that delivers results automatically, always-on application skills, and connector management from the headless CLI and SDK.

AIPOCH Open-Science is an open-source, local-first AI research workbench for scientists and researchers. It enables reproducible, inspectable research across models with scientific AI agents, Python and R execution, scientific data connectors, and cross-platform support for macOS, Windows, and Linux.

v0.27.0 grows the literature workspace into a dependable daily tool and lets long work step out of the way. The library imports many PDFs in one pass — selection, destination, duplicate policy, sequential progress, per-file results, stop, and retry — shows the active library's total at a glance, resolves DOIs extracted from imported PDFs into full metadata, and survives data-location relocation. Long Notebook, REPL, and shell work runs in the background: the agent turn is released while exact run identity, cancellation, and provenance are kept, and a delivery ledger hands completed results back automatically across local runs and remote compute jobs. Core application skills stay always enabled so built-in entry points never break, the headless CLI and Task SDK gain connector management, and headless Linux deployments get an explicit credential file store. Underneath it all, a stability pass of more than thirty fixes hardens the literature library end to end — cross-client sync, conflict-safe collection edits, import and metadata integrity, and export fidelity.

✨ Highlights

• Import PDFs in batches. The literature library imports many PDFs in one pass — selection, destination, duplicate policy, per-file progress, and retry of unfinished files — and the sidebar shows the active library's total at a glance. (#2379, #2286)
• Run long work in the background. Notebook, REPL, and shell jobs can run in the background: the agent turn is released while exact run identity, cancellation, and provenance are kept, and results are delivered automatically across local runs and remote compute jobs. (#2325)
• A broad literature stability pass. Library changes sync across clients immediately, stale collection edits surface a conflict instead of overwriting saved values, and data-location relocation preserves the library — alongside dozens of import, metadata, merge, citation, and export fixes. (#2368, #2366)
• Core skills stay available. Application skills that power built-in entry points — Customize, Environment & Packages, Remote Compute, and more — are always enabled, so those entry points can no longer break. (#2391)

🚀 New Features

• Batch PDF import — multi-select in the library's PDF import opens one dialog covering selection, destination, duplicate policy, sequential progress, per-file results, stop, and retry; single-file import keeps the metadata editor. (#2379, #2386)
• Background task execution — Python, R, persistent REPL, and shell work runs in the background with durable run identity, bounded shell concurrency, cancellation, and owned-process recovery, joined by a delivery ledger that suppresses duplicate delivery after direct observation. (#2325)
• Literature totals and richer PDF imports — the sidebar total updates with library changes independently of search and filters, and DOIs extracted from imported PDFs resolve to full metadata before the import draft opens. (#2286)
• CLI and SDK connector management — the headless CLI and Task SDK can list, inspect, and enable or disable built-in and custom research connectors.
• Always-on application skills — Customize, Environment & Packages, Remote Compute (SSH), Compute Environment Setup, Self-awareness, and Skill Creator are marked always enabled and can no longer be disabled into a broken state. (#2391)
• Credential file storage for headless Linux — Linux containers, servers, and WSL deployments without an OS credential vault can select an explicit file-storage mode at startup; OS storage stays the default. (#2387)

🔧 Improvements

• Mermaid diagram blocks animate height changes and offer a source/rendered toggle, so the transcript no longer jumps while diagrams load. (#2142)
• Library changes propagate to every open client immediately, and stale collection edits surface a retained-draft conflict with the latest saved values. (#2368)
• Data-location migration preserves and validates the literature library. (#2366)
• Windows update handoff explains installer failures, and multipart download errors are handled cleanly. (#2359, #2294)
• Blocked compute queue restoration explains the cause and offers retry. (#2384)

🐛 Bug Fixes

• Literature — a stability pass hardened the library end to end: verified PDF content identity (#2378); reviewed sources, saved metadata, and reading candidates (#2377, #2350, #2341); bounded attachment waits with recovered inbox undo and committed changes (#2375, #2351); protected chat history during deletion (#2374); confirmed attachment deletion with version history (#2373); visible import progress with cancellation (#2372); localized validation errors and merge labels (#2370); composition-safe editing and search (#2369); bounded large reads and checkpoint writes (#2367); deletion distinguished from cleanup and refresh failures (#2363); reliable save and search contracts (#2365); manual-entry metadata integrity (#2364); task recovery and metadata outcomes (#2354); aligned search and restore (#2353); inbox decisions and discovery provenance (#2335); validated batch capacity (#2352); validated PDFs with attachment recovery (#2337); recovered citation styles (#2349); full-text acquisition lifecycle and provenance (#2343); duplicate matches and member scope (#2347); complete export round trips (#2344); selection and replay identity (#2342); consistent collection workflows (#2338); note drafts and tag synchronization (#2340); review evidence integrity (#2345); import identities and author metadata (#2332); citation document and export fidelity (#2334); recovered failed operations with editing state (#2331); and literature dialogs raised above preview windows. (#2277, #2278)
• Notebook and compute — selected Windows R runtime paths are authorized (#2336); Windows kernel ownership persistence is restored (#2401); disabled interpreter probes are skipped during binding; agent file discovery stays within session scope (#2380); job inputs and recovery progress are preserved (#2279); analysis identity survives recovery (#2280); and concurrent compute-host profile and authentication updates no longer collide. (#2276)
• Previews and files — preview edits are preserved with resource recovery (#2392); single-page PDFs hide the reading entry (#2385); drafts survive restores and layout changes; JSON source and bounded text rendering are preserved (#2307); CSV integrity holds in bounded previews (#2298); PDF search ranges and navigation are corrected (#2297); TIFF sample meaning and page navigation recover (#2303); Office worksheets stay visible with session errors surfaced (#2304); Markdown keeps literal source with table keyboard actions (#2295); tab actions restore feedback and focus (#2284, #2305); incomplete script reconstructions are rejected; and artifact export selections and snapshots stay complete. (#2315)
• Sessions, projects, and composer — saved model preferences and provider defaults are restored (#2320); artifact bindings survive failed quits; project drafts and authoritative state are preserved (#2322); file-library selection and expanded ranges survive refresh (#2329); transcript reading windows and segment identities hold (#2327); composer drafts and referenced attachments survive mention selection (#2288, #2285); copied message references survive paste (#2402); send-now and queue actions validate correctly (#2289); cancelled task outcomes and run identities are preserved (#2319); elicitation keeps optional values and in-progress answers (#2317); Codex subscription side chats start and resume (#2394); and side-chat context holds through recovery. (#2272)
• Workspace, security, and platform — evidence and request error handling is hardened (#2330); local folder boundaries are enforced with browser-operation recovery (#2326); saved grants are clearer with the undo window preserved (#2269); annotation evidence and source navigation are restored (#2273); obsolete confirmations are released and dock windows restored (#2328); find overlays no longer block interactions (#2283); long dialog content stays reachable (#2299); workspace tooltips and archive authority stay in sync (#2268); background resizing is blocked behind preview modals (#2271); preview focus and keyboard ownership are preserved (#2323); and singular counts and metadata date locales are corrected. (#2324)

⚠️ Breaking Changes

• Remote-compute host details no longer include isSkeleton, and host instruction reads no longer synthesize resource summaries — the persisted instructions come back exactly as saved (empty before the first save), and custom consumers read the separate probe result for resources and pass the exact read document for guarded replacement. Saved instructions, SSH settings, authentication metadata, execution modes, and probe results are unchanged. (#2361)

📦 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, durable background execution for Notebook, REPL, and shell work with automatic result delivery across local runs and remote compute jobs, CLI and SDK connector management, branching into a new session from user messages or completed agent messages with persisted source lineage, composer session references (#) with turn-scoped read access, reversible archiving with keyboard undo, project pinning, a project quick switcher listing other active projects with title and description previews and fuzzy search once the list grows, persistent side conversations with advisories injected into running main turns, generated and editable session details, session hover previews in the sidebar, and SQLite-indexed summary-first session startup; in-app sandboxed previews for source links in agent responses; selectable Claude Code, OpenCode, Codex, and CodeBuddy agent frameworks (CodeBuddy app-managed and login-free) behind a shared provider turn-adapter interface; text, image, and PDF annotations that send selected context into conversations with click-to-reveal evidence, a session reading context that links up to three PDFs the agent can read, page through, and search, with agent configuration change markers in the timeline; opt-in persistent agent memory with project-scoped categories recalled across sessions and managed from Settings; production subagent delegation with durable messaging, restart recovery, structured output, and a per-session delegation switch; review-gated session plans with CLI plan controls; compact summary cards for artifact writes and notebook controls in messages and approvals; a unified composer lane with a session-scoped message queue, unified draft undo and redo history, and mid-turn Send now through native follow-up steering; hot-switching of compatible models and providers; multi-provider model configuration including Apodex, NVIDIA Build with a curated agent-capable catalog, the latest OpenAI and Anthropic model catalogs, Tencent Coding Plan and Token Plan subscription providers, an xAI OAuth subscription, a dedicated Vision model selector, custom token limits, and a consolidated Scenario models card; per-model-call usage details with a per-call context-window chart; a token usage dashboard with persisted per-run attribution; centralized credential management with guided recovery, device-wide shared credentials, and an explicit file-storage mode for headless Linux; a configurable reviewer model policy with an isolated review runtime and durable assessment snapshots; context-window composition insights with compaction boundaries; persistent Python/R/REPL kernels with bounded run-history payloads, live variable browsing, notebook and compute network access limited to user-approved domains with in-conversation approvals, safe managed-runtime reinstall, and a global toggle for agent-created runtime environments; session-scoped remote compute execution with a per-host execution mode (direct SSH or Slurm) with durable submission, polling, recovery, cancellation, and cleanup plus a guided Compute Environment Setup skill, with remote compute jobs that survive restarts and crashes through durable operation receipts and automatic recovery; a user terminal shared with the agent; app-managed and bring-your-own environments for Python and R; immutable, session-scoped artifact versions with checksummed content, producer code, execution history, exact input references, environment inventory, version-scoped reviewer evidence, and on-demand code reconstruction, with allowlisted text artifacts and uploads editable as raw text; a literature reference library with collections, project links, batch and identifier-aware imports, duplicate comparison and bulk merge, open-access full-text PDF attachment through Europe PMC, PMC, OpenAlex, arXiv, and Unpaywall with parallel multi-source lookup, citation formatting with artifact provenance, and relocation-safe storage; rich in-app previews for scientific data, documents, images, source code, molecular structures, and notebook history with right-click tab and content actions and full-screen mode; file attachments up to 10 GB with streaming upload; skills with conversational creation, import, marketplace browsing, explicit / selection, and always-on application skills; 24 built-in research connectors plus custom MCP servers; durable scoped permissions with allow grants, safe seeded defaults, and a restore-defaults action; remote-access pairing; interface localization in German, Spanish, French, Chinese (Simplified and Traditional), Japanese, Korean, and Russian; and auto-update with prominent update reminders and localized release notes.
• 🚧 Partial: R remains managed-only; provider choice remains constrained by the active framework's endpoint compatibility; remote compute covers direct SSH and Slurm (cloud-GPU submission is not built yet); skills remain local (no hosted public discovery commons); reproducibility stops at preserved evidence — captured file generations are groundwork, not yet automated reruns; and review is opt-in and record-scoped.
• 🗺️ Roadmap: a unified model gateway, a hosted public skills and specialist discovery commons, cloud-GPU execution, stronger sandboxing and credential isolation, and collaborative research workflows.

🐢 Known Limitations

• Remote compute covers direct SSH and Slurm. Cloud-GPU submission is not built yet.
• The literature library keeps maturing. Batch import, collections, duplicate merging, open-access PDF attachment, and citation formatting ship today; deeper reference-aware research workflows continue to evolve.
• Editable artifacts cover text formats. Markdown, plain text, scripts, and source code are editable as raw text; binary formats such as documents, images, and notebooks stay read-only, and editing always publishes new versions rather than rewriting history.
• Network sandboxing covers the app's notebook and compute runtimes, not the whole system. Processes you launch outside these runtimes are not subject to the domain allowlist. On Windows, the boundary applies only after the sandbox's one-time administrator setup; until then, notebook and compute code runs without it.
• Immutable file generations preserve earlier results, but deterministic reruns are still ahead. Captured generations are the groundwork; portable environment locks and full-fidelity session replay remain open.
• R is managed-only. A bring-your-own R interpreter path is 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.
• No hosted public specialist discovery commons. Specialist packages are portable across machines via import/export and the signed marketplace; what is not built yet is a hosted public discovery and forking hub.
• The task SDK is a first-generation surface. Task creation, polling, artifact retrieval, run progress, cancellation, session configuration, and connector management 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. Directory-level file access control is not built yet; notebook network sandboxing is the first enforced network boundary.
• 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, @VishvakR, @QiuLsG, @daanveer-tech, @wen2zhou, @roxi3906, and everyone in Discord, X, and Discussions.

---

Full Changelog: https://github.com/aipoch/open-science/commits/v0.27.0