AIPOCH Open-Science v1.0.4

AIPOCH Open-Science v1.0.4

A lighter, calmer workbench: CLI and desktop downloads shrink meaningfully, the standalone CLI exposes its research tools through a local MCP server, long conversations collapse intermediate messages and tool groups once turns settle, Literature references gain manual retraction and correction checks via Crossref, artifacts import from authorized folders, the GitHub Skills import workflow saves repositories, and question cards ship structured skip actions — with fixes across notebook, settings, session packages, and replay.

AIPOCH Open-Science is an open-source, local-first, model-agnostic 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.

v1.0.4 follows v1.0.3 with a payload-reduction and polish release. The packaging pass prunes unused generated components, bundles the translation worker against its compact runtime, ships a single bundled environment manager, and drops unused alternative distributions — while retaining every active runtime file and diagnostic, including the tokenizer lite runtime. The standalone CLI now speaks MCP, exposing its research tools to external MCP clients: the adapter connects to an already-running authenticated local profile service, acts under that profile's existing permissions rather than a separate sandbox or read-only view, and can persist writes or incur model costs. In the conversation surface, intermediate messages collapse after a turn settles and tool activity groups into summaries, with running, waiting, and interrupted work staying visible and search, annotations, and native find revealing hidden content on demand. Literature references with a DOI can be checked manually against Crossref for retractions and corrections, with observations persisted to the reference. Artifacts import from folders you authorize, the GitHub Skills import workflow can save repositories, and question cards offer structured skip actions on fixed, predictable controls.

✨ Highlights

• Downloads are meaningfully smaller. CLI and desktop runtime payloads are reduced by pruning unused generated components, bundling the translation worker against its compact runtime, shipping a single bundled environment manager, and dropping unused alternative distributions — without removing active runtime files or diagnostics. (#3863, #3867)
• The standalone CLI now speaks MCP. Research tools are exposed through a local MCP server for external MCP clients, under the selected profile's existing permissions — it is not a separate sandbox or a read-only view, so a connected client can persist writes or incur model costs. The profile's service must already be running and authenticated — start it with open-science start, then attach clients via open-science mcp; the adapter does not start the service itself. See the MCP CLI documentation for setup and the security guidance on treating MCP servers as untrusted until reviewed. (#3862)
• Long conversations read cleanly. Once a turn settles, intermediate messages collapse out of the way while running, waiting and interrupted work stays visible; tool groups collapse into summaries, and search, annotations, and native find still reveal hidden content when you need it. (#3868, #3866)
• Literature gains integrity checks. References with a DOI can be checked on demand against Crossref for retractions and corrections, with results and check times saved to the reference. A check sends the DOI to Crossref, and coverage is incomplete — finding no notices does not establish that a reference is unaffected. See the manual publication checks documentation for coverage and result semantics. (#3870)

🚀 New Features

• The standalone CLI exposes its research tools through a local MCP server for external tool clients. (#3862)
• Question cards support structured skip actions with fixed, predictable controls. (#3813)
• Literature references support explicit, per-reference retraction and correction checks via Crossref, started manually and only for entries with a valid DOI. (#3870)
• Artifacts can now be imported from folders you have authorized; the app reads the granted file and keeps a persistent immutable copy in its managed artifact storage. (#3882)
• The GitHub Skills import workflow can save repositories for reuse. (#3878)
• Conversation transcripts collapse intermediate messages after a turn settles, with a keyboard toggle and full reveal through search, annotations, and native find. (#3868)
• Tool activity in the transcript collapses into groups, with running tools summarized while they work. (#3866)

🔧 Improvements

• CLI and desktop downloads are smaller across platforms thanks to pruned runtime payloads, a compact bundled translation worker, and a single bundled environment manager. (#3863)

🐛 Bug Fixes

• Windows notebook standard execution is restored. (#3872)
• The tokenizer lite runtime is retained again in packaged builds. (#3867)
• Session package export no longer keeps waiting at the save-location step after a selection is made. (#3874)
• The settings folder access dialog and its cleanup step no longer race each other. (#3877)
• Ask-user selectors no longer loop after a continuation failure. (#3879)
• Replay presents automatic results as system events instead of ambiguous messages. (#3881)

📦 Install

Desktop client. Requirements: macOS 12+ (Apple Silicon or Intel), Linux x64 or ARM64, 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 updates itself in place.

Download the appropriate package from the Assets section below, or from the official download page. On macOS, the desktop app is also installable via Homebrew: brew install --cask open-science.

Platform	Package
macOS (Apple Silicon)	DMG for ARM64
macOS (Intel)	DMG for x64
Linux	AppImage or Debian package for x64 / ARM64
Windows	Code-signed installer for x64

Command line via Homebrew (macOS and Linux). The standalone CLI and local Web UI live in the official tap — bundled Node runtime, no separate Node.js needed, and upgrades through the normal brew flow:

brew install aipoch/tap/open-science
open-science start --no-open
open-science url

Command line via npm. With Node.js 22.13+, install the CLI (and Task SDK) globally; updates then follow your normal npm upgrade flow:

npm install --global @aipoch/open-science
open-science --help

Command line standalone (all platforms, no toolchain). The Assets below also include standalone CLI archives — macOS and Linux arm64/x64 tarballs and a Windows x64 zip — with a bundled Node runtime, so they install and run without Homebrew, npm, or a local Node.js. Linux packages additionally arrive as standalone .deb and .rpm installers.

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 notarized differently and may require approval through macOS Privacy & Security.

Windows — first launch. Official release builds are code-signed; Windows recognizes the publisher and no SmartScreen warning appears. 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: everything shipping in v1.0.3, plus the MCP server exposing the standalone CLI's research tools, conversation collapsing for intermediate messages and tool groups, manual Crossref retraction and correction checks in Literature, artifact imports from authorized folders, saved repositories in the GitHub Skills import workflow, structured skip actions on question cards, the reduced CLI and desktop payloads, and the fixes described above.
• 🚧 Partial: 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); WSL2 Bash remains an opt-in Windows x64 preview; the skills commons ships marketplace discovery but not cross-machine forking or user-facing version pinning; reproducibility ships replayable per-version verification, session replay, and RO-Crate packaging, while solver-exact equivalence remains open; review is opt-in and record-scoped.
• 🗺️ Roadmap: a unified model gateway, cross-machine skill forking and version pinning, cloud-GPU execution, stronger sandboxing and credential isolation, and collaborative research workflows.

🐢 Known Limitations

• Replay reproduces recorded steps, it does not re-execute them. Solver-exact equivalence guarantees remain open; replayable verification remains the deterministic check, and starting a check from the web client currently requires the desktop app.
• Imported .science sessions are read-only. They stay inspectable and referenceable, and forking them into a writable continuation remains available; execution and continuation remain disabled on the import itself.
• PDF annotations are bound to the file version they annotate. Notes follow the file across projects and sessions where that version is shared, but they do not sync across machines.
• Bookmarks belong to a single session. They do not follow branch switches, join .science package exports, or sync across machines — forking copies them into the new session with fresh identities.
• WSL2 Bash is a preview. It is opt-in on Windows x64 only; PowerShell remains the default shell runtime.
• Remote compute covers direct SSH and Slurm. Cloud-GPU submission is not built yet.
• Editable artifacts cover text formats. Binary formats stay read-only, and editing always publishes new versions.
• Network sandboxing covers the app's notebook and compute runtimes, not the whole system. On Windows, the boundary applies only after the sandbox's one-time administrator setup.
• Provider choice is per framework, not one unified gateway. The available protocol depends on the selected agent backend, and remote endpoints must use HTTPS.
• Code reconstruction is LLM-generated. It does not replace deterministic reproduction; replayable verification is the deterministic check.
• The reviewer is opt-in and record-scoped. It does not replace domain-specific validation of citations, units, statistics, or methods.
• No local GPU compute backend.
• No multi-user real-time collaboration.

🙏 Acknowledgements

Thanks to @ewen-poch, @nasus2002, @roxi3906, @daanveer-tech and everyone in Discord, X, and Discussions.

---

Full Changelog: https://github.com/aipoch/open-science/commits/v1.0.4