AIPOCH Open-Science v1.0.6

AIPOCH Open-Science v1.0.6

Zotero libraries import into the literature library, PDFs read in automatic order, Windows installation gets a third faster while preserving co-located research data, artifacts stay publishable through long-running turns, and OpenCode sessions, Windows startup, and Slurm outcome recovery are fixed across platforms.

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.

v1.0.6 follows v1.0.5 with a reliability-and-migration release. The literature library now imports a personal Zotero library or selected collections from Zotero 10 or later on the same computer, carrying metadata, PDFs, tags, and collection hierarchies without duplicating existing references, and reference search covers personal notes. Linked PDFs drop their persistent reading-order configuration in favor of automatic ordering that prefers the document's own logical tags. On Windows, installation drops from a ~21 s median to ~14 s with a third fewer payload files, updates no longer delete research data stored next to the application, and deep install paths complete reliably. Artifact capabilities now live for the entire open assistant turn, so long-running sessions keep publishing. OpenCode sessions gain three fixes — tool calls with late-arriving names, hidden 1.x compaction summaries, and Windows launch limits for large configurations.

✨ Highlights

• Import your Zotero library. The literature library gains Import from Zotero: bring a personal library or selected collections from Zotero 10 or later on the same computer, carrying bibliographic metadata, PDFs, tags, collection hierarchies, and source links, with existing references matched instead of duplicated. (#3949)
• PDF reading order is automatic. Opening a linked PDF no longer shows persistent reading-order controls above the document. Extraction prefers the file's own logical tags and falls back to layout analysis when tags are missing or ambiguous, and historical messages with manual regions read in the new automatic order too. (#3979)
• Windows setup is faster and safer. Installation drops from a ~21 s median to ~14 s with a third fewer payload files and parallel extraction, updates no longer delete research data stored next to the application, and installs to deep folder paths complete instead of failing at the copy step. (#3957, #3963, #3946)
• Long-running turns keep their artifacts. Artifact publication no longer expires two hours into a still-running assistant turn: capabilities now live for the entire open turn, so overnight and long-batch sessions publish their files as before. (#3980)

🚀 New Features

• Search the literature library by personal notes — a note like "this method fits my data" now finds its reference through the normal search box. (#3942)
• Custom stdio connectors configured with npx or npm now run on machines without those commands, using the application's bundled Node and npm runtime while installed commands keep priority. (#3954)

🔧 Improvements

• Translated PDFs open faster: saved-translation source verification now overlaps reads in small batches on first open, and the unused local translation worker was removed from the installation, shrinking its size. (#3982, #3978)
• Slurm queue, controller, and accounting observations resolve independently; terminal job outcomes are persisted before log collection, cancellations deliver exactly one final result, and short Windows SSH commands no longer hang on an idle input pipe. (#3937)
• Split-button menus — conversation send/queue/stop, permission approvals, literature conversation selection, PDF marking and area tools — share one connected panel with automatic placement, reversible animations, and independently usable primary and arrow buttons. (#3925)
• Reported security alerts are addressed with compatible dependency updates and bounded parsing in place of backtracking-prone pattern matching. (#3956)
• WSL setup guidance now states up front that mirrored networking requires Windows 11 22H2 or later, instead of advising a mode Windows 10 cannot supply. (#3945)
• Interactive credential authorization waits up to five minutes for your approval, and timeout recovery explains what happened instead of looking like a missing-key problem. (#3939)

🐛 Bug Fixes

• OpenCode sessions no longer end with an internal error when a tool call's name arrives after its first output fragment, 1.x compaction summaries stay out of conversations and history replay, and large configurations no longer fail to launch on Windows. (#3970, #3967, #3948)
• Windows startup accepts cookie stores with binary payload fields instead of blocking the launch preflight with a decryption failure. (#3955)
• Native Windows R sessions work when the application's temporary directory contains spaces. (#3958)
• The desktop window no longer loses synchronization when a session snapshot read arrives while all large-transfer channels are busy. (#3969)
• Sending a first message no longer shows thinking and elapsed time before you actually submit, and stopping preparation restores your unsent draft. (#3931)
• Imported sessions show past delivery failures as collapsed, expandable history instead of a current error above the fork action. (#3941)
• The DOCX preview toolbar stays above document content while scrolling, at any zoom. (#3743)

📦 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 v1.0.5, plus Zotero library import and personal-note reference search in the literature library, automatic PDF reading order, faster and data-preserving Windows installation with long-path support, turn-scoped artifact capabilities, bundled npm runtime fallback for custom connectors, 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, @wen2zhou, @roxi3906, @nasus2002, @daanveer-tech and everyone in Discord, X, and Discussions.

---

Full Changelog: https://github.com/aipoch/open-science/commits/v1.0.6