Repository navigation
Releases
Research Pilot 0.8.8
Research Pilot 0.8.8
This release includes the changes since public version 0.8.6.
What's new
- Teams in PIAgent conversations. Discuss a goal, review the brief card, and click Start team. Follow messages, questions, member activity and results in the conversation. New teams coordinate through tasks and direct messages without a mandatory research graph or fixed author hierarchy. Research-to-manuscript work still requires a separate writing brief and user-confirmed launch.
- Browser-assisted research and PDF import. PIAgent can help with browser tasks and paper full-text retrieval. Downloading a file does not automatically import it into the library. Desktop computer-use support is macOS-only.
- GitHub Copilot support. Sign in under Settings → Subscriptions and use supported models in PIAgent. The same settings area provides Claude Code and Codex setup actions. Model access and quotas depend on the provider and account.
- Reusable domain knowledge under user control. Review rule candidates, approve content versions, choose which rules to use, and delete unwanted rules. Opening the review panel clears the new-rule notification without enabling rules.
- Team context and recovery improvements. Updated state delivery and compaction handling, task authorization, required-deliverable checks, publication recovery and final checkpoints.
- Workbench, history and phone refinements. Cached session-search text, mathematical notation in chat, file-link navigation fixes, composer scrolling improvements and phone rendering updates.
- Updated runtime. Electron 44.11.3, a terminal WebGL font-sizing fix, and selected fixes in the maintained PI runtime for streamed tool arguments and Anthropic error results containing images.
Downloads
| Platform | Packages |
|---|---|
| macOS Apple Silicon | Signed, notarized .dmg and .zip |
| Linux x86_64 | .deb and .AppImage |
| Linux arm64 | .deb and .AppImage |
Windows and Intel Mac installers are not included. GitHub's automatic source archives contain the public documentation repository, not the application installer or private application source.
Release checks
- Local release gate passed: 3,225 tests passed, 8 skipped, no failures; Electron typecheck, production builds, packaging budgets and the end-to-end demo passed.
- Terminal glyph sizing, startup UI, desktop/phone team cards, domain-knowledge review and CLI settings checks passed with isolated fixtures and no live model calls.
- macOS code signatures, stapled notarization tickets and Gatekeeper assessments passed. The mounted DMG and extracted ZIP passed bundled-runtime checks for PI imports, SQLite, PTY, daemon authentication and MCP search. The mounted DMG also passed the packaged startup-UI check.
- On Ubuntu 24.04 containers for arm64 and x86_64, installed deb packages and extracted AppImages passed checks for PI imports, native SQLite and PTY, bundled daemon authentication, MCP search, startup UI and the preload bridge.
- These automated Linux checks use a virtual display and
--no-sandbox; x86_64 runs under emulation. The x86_64 AppImage payload was extracted with SquashFS tools, not its bootstrap. These checks do not establish native-desktop, AppImage launcher/FUSE-mount or sandbox compatibility on every distribution. - One repeated gate run hit a process-cleanup assertion. The targeted test, Compute suite and subsequent full gate passed without code changes.
- GitHub CI was manually disabled in the source repository; this release does not claim a GitHub CI pass. Live subscription-provider calls are not part of these checks.
Upgrade and installation
Finish active work and quit Research Pilot before replacing the app. Reopen it after installation so the app and bundled services use the same version. Keep backups of ~/.pipilot and your project folders. Do not delete research data to upgrade.
On macOS, open the DMG and drag the app into Applications. On Ubuntu/Debian, prefer the .deb package. AppImages may need libfuse2t64 on Ubuntu 24.04 (libfuse2 on older releases); arm64 AppImages also need zlib1g-dev. An AppImage may need --no-sandbox on Ubuntu 24.04. Linux team shell execution requires working bubblewrap isolation; execution is refused when isolation cannot be established.
Permissions
Desktop and phone sessions launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents can run commands, edit or delete files, and use the network with your OS account's permissions without per-action approval. This does not grant administrator privileges. Use trusted tasks and keep backups.
Research Pilot 0.8.6
Research Pilot 0.8.6
What's new
- Files and conversations in one workbench. Open files beside PI Agent conversations without switching to a separate top-level Files tab. Use left/right split panes to keep files and conversations side by side.
- More responsive long conversations. Long chats use a windowed message list, inactive conversation feeds avoid unnecessary updates, and cached viewers reduce repeat work when switching tabs.
- Context-window meter. A thin bar shows context usage relative to the compaction threshold, helping you decide when to start a new conversation.
- Improved file viewers. Updates cover PDF search and text selection, image viewing, CSV and notebook content copying, and file navigation.
- PI recovery improvements. Refined handling for transient connection errors and compaction failures.
Downloads
Six installation packages are provided:
| Platform | Packages |
|---|---|
| macOS Apple Silicon | Signed, notarized .dmg and .zip |
| Linux x86_64 | .deb and .AppImage |
| Linux arm64 | .deb and .AppImage |
Windows and Intel Mac installers are not included. GitHub's automatic source archives contain the public documentation repository, not the application installer or private application source.
Release checks
- Release gate passed: 2,768 tests passed, 8 skipped, no failures; Electron typecheck, production builds, packaging budgets and the end-to-end demo passed.
- macOS code signatures, stapled notarization tickets and Gatekeeper assessments passed. Both the mounted DMG and extracted ZIP passed packaged-runtime smoke checks.
- On Ubuntu 24.04 containers for arm64 and x86_64, both installed deb packages and extracted AppImages passed checks for startup, bundled daemon and MCP search, native SQLite and PTY, and a PI turn against a local scripted model.
- Split-pane interaction and workbench performance checks passed in Electron with fixture IPC and no live model calls.
- These automated Linux checks use a virtual display and
--no-sandbox; they do not establish native desktop, FUSE-mount or sandbox compatibility on every Linux distribution. Live Claude/Codex subscription calls were not part of the release checks.
Upgrade
Finish active work and quit Research Pilot before replacing the app. Reopen it after installation so the app and bundled services use the same version. Keep backups of ~/.pipilot and your project folders. Do not delete research data to upgrade.
On Ubuntu/Debian, prefer the .deb package. AppImages may need libfuse2t64 on Ubuntu 24.04 (libfuse2 on older releases); arm64 AppImages also need zlib1g-dev. On Ubuntu 24.04, an AppImage may need --no-sandbox.
Permissions
Desktop and phone sessions launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents can run commands, edit or delete files, and use the network with your OS account's permissions without per-action approval. This does not grant administrator privileges. Use trusted tasks and keep backups.
Research Pilot 0.8.5
Research Pilot 0.8.5
This release adds guided setup, side-by-side files and PI conversations, editable paper diagrams, and a separate host for local work.
What's changed
- Guided setup. The welcome panel and Settings → Setup explain missing authentication and tools. Required and optional components are grouped separately, and supported keys in shell configuration can be recognized without entering them again. Features prompt for their own requirements when needed.
- Multiple PI conversations and split panes. Open several PI Agent tabs, reopen saved conversations from Recap, and split Files or PI Agent into left and right panes. PI Agent appears immediately after Files in the project navigation.
- Paper diagrams with draw.io. Edit
.drawiofiles inside Files, export figures, and ask the figure assistant to revise the canvas. Final outputs remain visible; supporting figure files live in.drawio-figure-work. - A separate local host. Local terminals, PI sessions, teams, and the phone gateway run outside the desktop window. Desktop restarts can keep that work running. Public installers start their bundled host and do not need a checkout or a separate host installation. File → Close Project and Stop Its Work helps release a project folder before unmounting its disk.
- PI tools and long sessions. Background shell commands can be checked and stopped. Tool output paging and stored results preserve access to large outputs, and smaller UI event payloads reduce pressure on event streams. Context handling includes revised compaction budgets and provider-overflow recovery.
- Files and reading. The file tree follows the active file, opening a file reuses its existing tab, and PDF text can be selected and copied. Terminal links open in the system browser. Image detection uses file contents.
- A simpler interface. Recap, Literature, Research, and Team views have layout and navigation improvements. The retired Memory settings page has been removed; project-memory loading, retrieval, and agent-directed saving remain available. The legacy provenance checker has also been removed for now.
- Phone and model controls. The phone can select among PI Agent tabs. PI replies identify their model and reasoning effort, and the model picker includes GPT-6.1 Terra where the configured provider offers it.
Downloads and installation
Choose a named application package below. The automatic “Source code” archives contain this public documentation repository, not the application source or an installer.
| Platform | Recommended | Alternative |
|---|---|---|
| macOS Apple Silicon | Research-Pilot-0.8.5-mac-arm64.dmg |
Research-Pilot-0.8.5-mac-arm64.zip |
| Linux x86_64 | Research-Pilot-0.8.5-linux-amd64.deb |
Research-Pilot-0.8.5-linux-x86_64.AppImage |
| Linux arm64 | Research-Pilot-0.8.5-linux-arm64.deb |
Research-Pilot-0.8.5-linux-arm64.AppImage |
- macOS: open the DMG and drag Research Pilot into Applications.
- Ubuntu/Debian: install the deb with
sudo apt install ./Research-Pilot-0.8.5-linux-amd64.deb, substitutingarm64on ARM. - AppImage: make the file executable. Ubuntu 24.04 needs
libfuse2t64; older Ubuntu versions uselibfuse2. The arm64 AppImage also needszlib1g-dev. Prefer the deb if Chromium's sandbox prevents startup;--no-sandboxis a fallback that disables that sandbox for the launch. - Agent setup: Claude and Codex tabs use the corresponding installed and authenticated CLI. PI Agent's runtime is bundled; configure its model and authentication in Settings. Packaged use does not need Node.js, Docker, or the source repository.
See the installation guide.
Upgrade and permissions
Finish active work, back up ~/.pipilot and your project folders, then quit the old app with local work stopped before replacing it. Reopen the new application and start a new agent session to receive the updated instructions and skills. Refresh the phone companion after upgrading. Do not delete research data to upgrade. Keep PI checkpoint files with their sibling image directories.
Desktop and phone sessions launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents can edit or delete files, run commands, and access the network with your OS account's permissions without per-action approval. Codex's own sandbox is disabled; this does not grant administrator privileges. This is fixed launch behavior. Use trusted tasks and keep backups.
Verification
-
The final source passed the local
npm run verify:releasegate: 2,709 tests passed, 8 skipped, 0 failed. Electron typechecking, production builds, release size/dependency budgets, and the end-to-end daemon/MCP demo passed. The test command now selects the project's tests and does not discover ignored third-party checkouts inanalysis/. -
macOS code signatures, Apple notarization, stapled tickets, and Gatekeeper assessments passed. The application was tested from the mounted DMG and the extracted ZIP with isolated profiles.
-
On Ubuntu 24.04, installed DEBs and extracted AppImage payloads for arm64 and x86_64 passed checks for application startup with an empty home directory and no separately installed host, PI runtime imports, native SQLite and PTY support, daemon version/protocol, authenticated Teams access, unauthenticated-access rejection, MCP initialization and searches, renderer-to-host access, and a PI turn against a local scripted model. The installed DEB sandbox helpers have root ownership and mode 4755.
-
All six package payloads report version 0.8.5, source commit
01b2f06fa589188e8edca280988cd706987cbc64, and the same clean build fingerprint. The startup UI, headless host, and ASAR integrity fixture checks also passed.analysis/is not tracked in the source repository and is absent from the packaged app and resources. -
All six uploaded installers match the tested local files by size and SHA-256.
Verification limits: checks were local. Linux tests used containers, virtual displays, and --no-sandbox; x86_64 ran under emulation on Apple Silicon. The x86_64 AppImage's bootstrap extraction failed under emulation, so its filesystem was extracted with unsquashfs before testing. These checks do not establish native x86_64 launcher, desktop sandbox, or FUSE-mount behavior. The existing macOS installation was not replaced. No live-model or physical-phone tests were performed.
SHA-256 checksums
698c716dc621cd7700b93afefac0ecb57e2648a560fd08f4d144246713098f35 Research-Pilot-0.8.5-linux-amd64.deb
6253adb109e68181dad49ca38aa75f78a1448b0783989af3b3dc84ba27a5719e Research-Pilot-0.8.5-linux-arm64.AppImage
ab80bc2d9b07241e75b19213de1792013d25ebbe60e61f54219a14676bbb0385 Research-Pilot-0.8.5-linux-arm64.deb
4af5f0b3882295b9c4e97c08ccaeb25e7fe91ec5dec6712465c3a00ae1a47a7b Research-Pilot-0.8.5-linux-x86_64.AppImage
6be300121e8af0e2b884f121ff093faa1e378e4777a05e199bdedbb98efeb3e2 Research-Pilot-0.8.5-mac-arm64.dmg
fb6e6d280d716d97160ef0c06abbdac4fcff0ffb87ec6ad2d0fb63397162eb74 Research-Pilot-0.8.5-mac-arm64.zip
Research Pilot 0.8.4
Research Pilot 0.8.4
A bug-fix release focused on long PI sessions, saved images, subscription model identity, and reliable launch instructions.
What's changed
- An unusable compaction result no longer ends an otherwise usable PI turn. Empty, truncated, or non-shrinking summaries leave the previous working context intact. PI continues if that context still fits the input budget; if it does not fit, PI reports the capacity problem without replacing the original history. Authentication, transport, cancellation, and persistence errors are still errors, not silently ignored failures.
- PI session images are stored separately from checkpoint JSON. Conversation history, working context, and pending inputs share image files and restore their original encoding when loaded. Missing or corrupt image data is detected rather than sent to the model as an unresolved reference. This avoids repeatedly embedding the same image data in large checkpoints.
- Subscription routes retain model-specific behavior. Codex context windows are selected per model rather than uniformly using the channel fallback. Anthropic subscription authentication is separated from the SDK model identity, so restored signed thinking blocks keep the provider identity expected by the SDK.
- PiPilot identity and rules travel with the session launch. Claude and Codex no longer depend on PiPilot-written project guidance for those instructions. Workspace sync retires older PiPilot-managed blocks in
AGENTS.mdandCLAUDE.md, preserving user content. A separately launched CLI does not receive PiPilot's launch identity or rules; shared skills and delegate definitions remain visible by design. - Builder instructions now ask for evidence before completion claims. Agents are asked to check claims against results from the current session after the relevant edits and state what remains unverified. This is model guidance, not an enforced guarantee of correctness.
- UI verification scripts were brought back into line with the current interface, with documented checks for each feature area.
Downloads and installation
Download a named package below, not GitHub's automatic “Source code” archives. This public repository distributes application packages and documentation, not the application source.
| Platform | Recommended | Alternative |
|---|---|---|
| macOS Apple Silicon | Research-Pilot-0.8.4-mac-arm64.dmg |
Research-Pilot-0.8.4-mac-arm64.zip |
| Linux x86_64 | Research-Pilot-0.8.4-linux-amd64.deb |
Research-Pilot-0.8.4-linux-x86_64.AppImage |
| Linux arm64 | Research-Pilot-0.8.4-linux-arm64.deb |
Research-Pilot-0.8.4-linux-arm64.AppImage |
- macOS: open the DMG and drag Research Pilot into Applications. The application and DMG are signed, Apple-notarized, and stapled.
- Ubuntu/Debian: install the deb with
sudo apt install ./Research-Pilot-0.8.4-linux-amd64.deb, substitutingarm64on ARM. The deb is recommended. - AppImage: make the file executable. Ubuntu 24.04 needs
libfuse2t64; older Ubuntu versions uselibfuse2. The arm64 AppImage also needszlib1g-dev. If Chromium's sandbox prevents startup, prefer the deb;--no-sandboxis a fallback that disables that sandbox for the launch. - Agent setup: install and sign in to Claude Code and/or Codex CLI for those tabs. The built-in PI Agent does not need a separate CLI; configure its model and authentication in Settings. Packaged use does not need Node.js or Docker.
See the installation guide.
Upgrade and permissions
Finish active work, back up the entire ~/.pipilot directory and your project folders, quit the old app, replace it, and reopen. Start a new agent session to receive the new launch instructions. Refresh the phone companion after upgrading. Do not delete research data to upgrade. Keep PI checkpoint files and their sibling image directories together; older releases cannot read the new image-reference checkpoint format.
Desktop and phone sessions launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents can edit or delete files, run commands, and access the network with your OS account's permissions without per-action approval. Codex's own sandbox is disabled; this does not grant administrator privileges. This is fixed launch behavior, not a selectable mode. Use trusted tasks and keep backups.
Verification
- Local release gate, Node.js 20.20.2: the final source passed
npm run verify:releasetwice. Each run reported 2,456 tests passed, 8 skipped, and 0 failed or cancelled. Electron typechecking, production builds, release size/dependency checks, and the end-to-end MCP/daemon demo passed. Initial runs exposed two test timing issues; the fixtures now wait for a stopped Compute worker to exit and keep periodic heartbeats from racing the lease-expiry test's simulated clock. These changes do not alter production behavior. - macOS Apple Silicon: code signatures, Apple notarization, stapled tickets, and Gatekeeper assessments passed. The app launched from the mounted DMG with an isolated profile. Both DMG and ZIP payloads passed PI runtime imports, native SQLite and PTY checks, daemon health/protocol checks, authenticated Teams API access, rejection of unauthenticated access, and packaged MCP initialization, Wiki search, and knowledge search.
- Ubuntu 24.04 arm64 and x86_64: installed deb packages and extracted AppImage payloads passed start-screen, PI imports, SQLite, native PTY, authenticated daemon/Teams, and packaged MCP search checks. The installed deb sandbox helper has root ownership and mode 4755.
- Package consistency: all three targets report version 0.8.4. Their daemon, MCP, Builder instructions, and phone JavaScript bytes match. Builder revision 8 is present in each package. The ASAR integrity check also passed, including detection of a modified archive. All six uploaded assets match the tested local files by size and SHA-256.
Limits: GitHub-hosted CI did not start because of the account billing/spending limit; it is not a passing CI run. Linux checks used containers and a virtual display with --no-sandbox; x86_64 ran under emulation on Apple Silicon. The x86_64 AppImage payload was extracted with unsquashfs, not its bootstrap. These checks do not establish native x86_64 launcher, desktop sandbox, or FUSE-mount behavior. No live model calls or physical-phone tests were made for this release. The installed macOS application was not replaced, and scripted process cleanup is not a graceful-quit test.
SHA-256 checksums
6379bdf6b881fb24fb3786917217757400a6b3f83101bc1bd43ff9263f245574 Research-Pilot-0.8.4-linux-amd64.deb
d15fa45c829eaf2ed856e667d152b044abdebcd354be43a097748860c1972798 Research-Pilot-0.8.4-linux-arm64.AppImage
cc0fca8463e5b611e2aee9c6a2d83f84e0b3bb984f83a7c04b85085fd517c843 Research-Pilot-0.8.4-linux-arm64.deb
4d9e01fc169d3897a37bbeaf6b4d24edcd9999955f38fdefb08b912f2164516a Research-Pilot-0.8.4-linux-x86_64.AppImage
66bd99520c0544a87056fcc1317d9f7fff29e44a0c2aeb44f549ecf82ecab2a4 Research-Pilot-0.8.4-mac-arm64.dmg
48655161d96a770fb07ee55c1093775f772b6e958ea4d172aeb028ce8a8be30a Research-Pilot-0.8.4-mac-arm64.zip
Research Pilot 0.8.3
Research Pilot 0.8.3
This release improves PI conversations, context handling, tool consistency, and remote-file workflows. The README and installation guide have been updated for 0.8.3.
What's changed
- PI conversations follow the order of work. Markdown renders while replies stream. Explanations and tool calls appear in sequence rather than being collected out of order. Text between tool calls has a lighter style on desktop and phone, so the final answer remains distinct. Copying or saving an answer excludes those intermediate explanations.
- Context handling uses provider counts when available. PI retains the provider-reported token baseline across history changes and estimates the change since that baseline. A context-overflow rejection triggers one compaction-and-retry attempt. Token accounting, cache affinity, and persisted usage have also been corrected.
- Tools share more of their implementation. PI and MCP tools use shared definitions and explicit input/output contracts. Follow-up fixes address integration gaps without making each agent's available tool set identical.
- Remote files open in the right workspace. HTML previews opened from chat preserve the remote workspace URI. Remote downloads can replace a file when the save dialog has confirmed the replacement.
- Work modes are simpler. Built-in modes replace editable custom mode configurations; new sessions use the current built-in version. Changing modes still requires confirmation when foreground sessions must end.
- Team progress is less repetitive. Member progress notices are batched for the lead, and a task that repeatedly loses its host is stopped rather than retried indefinitely.
- Upgrades can restart the previous daemon protocol. The restart path remains authenticated.
Downloads and installation
Download the named package for your platform, not GitHub's automatic “Source code” archives. This public repository distributes application packages and documentation, not the application source.
| Platform | Recommended | Alternative |
|---|---|---|
| macOS Apple Silicon | Research-Pilot-0.8.3-mac-arm64.dmg |
Research-Pilot-0.8.3-mac-arm64.zip |
| Linux x86_64 | Research-Pilot-0.8.3-linux-amd64.deb |
Research-Pilot-0.8.3-linux-x86_64.AppImage |
| Linux arm64 | Research-Pilot-0.8.3-linux-arm64.deb |
Research-Pilot-0.8.3-linux-arm64.AppImage |
- macOS: open the DMG and drag Research Pilot into Applications. The app and DMG are signed and Apple-notarized.
- Ubuntu/Debian: install the deb with
sudo apt install ./Research-Pilot-0.8.3-linux-amd64.deb, substitutingarm64on ARM. The deb is recommended. - AppImage: make the file executable. Ubuntu 24.04 needs
libfuse2t64; older Ubuntu versions uselibfuse2. The arm64 AppImage also needszlib1g-dev. If Chromium's sandbox prevents startup, prefer the deb;--no-sandboxis a fallback that disables that sandbox for the launch. - Agent setup: install and sign in to Claude Code and/or Codex CLI for those tabs. The built-in PI Agent does not need a separate CLI; configure its model and authentication in Settings. The application includes its runtime, so packaged use does not need Node.js or Docker.
Upgrade and permissions
Finish active work, back up ~/.pipilot and your project folders, quit the old app, replace it, and reopen. Refresh the phone companion after upgrading. Do not delete research data to upgrade. Existing custom work modes are no longer editable in this version.
Desktop and phone sessions launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents can edit or delete files, run commands, and access the network with your OS account's permissions without per-action approval. Codex's own sandbox is disabled; this does not grant administrator privileges. This is fixed launch behavior, not a selectable mode. Use trusted tasks and keep backups.
Verification
- Local release gate, Node.js 20: 2,426 tests passed, 8 skipped, 0 failed or cancelled. Electron typechecking, production builds, release size/dependency checks, and the end-to-end MCP/daemon demo passed. A test-lifetime issue found during verification was repaired: the rate-limiter concurrency test now keeps its event loop alive while requests are queued; the limiter's production behavior is unchanged.
- macOS Apple Silicon: code-signature, Apple notarization, stapled-ticket and Gatekeeper checks passed. The application launched from the mounted DMG with an isolated profile. Both DMG and ZIP payloads passed PI runtime imports, native SQLite and PTY checks, daemon health/protocol checks, authenticated Teams API access, rejection of unauthenticated access, and packaged MCP initialization, Wiki search and knowledge search.
- Ubuntu 24.04 arm64 and x86_64: installed deb packages and extracted AppImage payloads passed application-start-screen, PI imports, SQLite, native PTY, authenticated daemon/Teams and packaged MCP search checks. The deb sandbox helper has root ownership and mode 4755.
- Phone page: deterministic UI checks passed for intermediate text, expandable thinking and tool details, live progress, final-answer uniqueness, reload and pagination. This is not a physical-phone or live-model test.
- Package consistency: versions and packaged layout were checked on all three targets. The daemon, MCP and phone JavaScript bytes match across platforms. All six uploaded assets match the tested local files by size and SHA-256.
Limits: GitHub-hosted CI did not run; the release gate ran locally. Linux checks used containers and a virtual display, with --no-sandbox for the GUI checks; x86_64 ran under emulation on Apple Silicon. The x86_64 AppImage payload was extracted with unsquashfs rather than running its bootstrap. These checks do not establish native x86_64 launcher, desktop sandbox or FUSE-mount behavior. Packaged smoke tests made no live model calls, and the existing installed macOS application was not replaced.
SHA-256 checksums
8ee4f7de3cd3226689dc72386a5b911d14532d1137821596526e3caa4a39efeb Research-Pilot-0.8.3-linux-amd64.deb
8dc9885978e6321da4274115852d4c2cb97fbdc6edbe6ae1ed71156947977d67 Research-Pilot-0.8.3-linux-arm64.AppImage
84917adfe776663ada4a22a87a22daac12d2c4dc899631e046cfb1528bc8cbb4 Research-Pilot-0.8.3-linux-arm64.deb
33ce7dc68e2baac38c8e4778facd47c004dc626a4a08a6e61be35974fe8f1f9a Research-Pilot-0.8.3-linux-x86_64.AppImage
23822c3f63b66a3a33c226e12148fcf94db80295091a258983a33ab687fe7ab2 Research-Pilot-0.8.3-mac-arm64.dmg
f6d8000a5d00ccf074b5137cf4ec28a80565b6bfe4462317f979b1f818752c23 Research-Pilot-0.8.3-mac-arm64.zip
Research Pilot 0.8.2
Research Pilot 0.8.2 adds connected workspaces across computers, brings team conversations and research progress to the phone, and improves memory, startup responsiveness and recovery during long research runs.
What changed
- Work across computers. Connect computers through Settings → Servers, browse projects grouped by machine, and open remote sessions and files. Connection recovery preserves existing work; remote file actions and HTML previews now use the correct machine and paths.
- Project memory with sources and history. Agents save reusable preferences, configuration references and observed experience through the same Artifact tools used for research material. Ordinary sessions, including Builder, receive compact project-memory summaries by default. Records keep their source context and prior versions; users can correct or forget them.
- Team conversations and continued rounds. Research and Manuscript Teams have an assistant drawer for questions and updates. Assistants pass on the user's recorded words. Continuing a finished team creates a linked round that shares the working folder and preserves the previous round. Checkpoint recovery, blocked status, Compute submission recovery and final coordinator state have been improved.
- Resident research and phone controls. Resident researchers support assistant conversations, HTML briefings, clearer progress reports and improved scheduling and recovery. The phone includes the Team board and assistant drawers, resident controls, and individual device pairing and revocation.
- More responsive long-running work. Research records and histories use incremental storage and paginated reads. Startup, knowledge search and Wiki indexing avoid several blocking operations. Persistent status indicators use still or stepped states to reduce continuous rendering.
- More reliable agent sessions. Fixes cover expired subscription authentication, retry progress, PI session identity after reconnects, reply duration, and Codex MCP identification of the current conversation. File and image previews and terminal theme handling have also been refined.
- Research and writing guidance. Updated skills emphasize the strongest contribution relative to the nearest work, preserve claim strength during wording edits, and keep reports focused on evidence and decisions.
Installation and permissions
macOS Apple Silicon: open the DMG and drag Research Pilot into Applications.
Ubuntu/Debian: install the matching deb with sudo apt install ./Research-Pilot-0.8.2-linux-amd64.deb (use linux-arm64.deb on ARM). The deb is recommended. AppImage users on Ubuntu 24.04 may need libfuse2t64 and --no-sandbox; ARM AppImages also need zlib1g-dev. No Windows or Intel Mac installer is included.
Install and sign in to Claude Code and/or Codex CLI before using their tabs. PI Agent's runtime is bundled; configure its model and authentication in Settings. Desktop and mobile launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents act with your user account's filesystem and network permissions. These flags do not grant administrator rights.
Finish active work and quit the existing application before replacing it so background processes load the new version. Research data and settings remain under ~/.pipilot; back up important data before upgrading. Update connected computers as well.
Verification
-
Node 18 and Node 20: 2,312 tests passed and 8 skipped on each version. The release gate also passed desktop type checking, production builds, size/dependency checks and the isolated MCP/daemon demo.
-
macOS arm64: the app and DMG passed Apple notarization and Gatekeeper; stapled tickets were verified, including the ZIP's app. The mounted DMG and extracted ZIP both passed packaged workspace-screen, PI imports, SQLite, native PTY, authenticated daemon/Teams and MCP Wiki-search checks.
-
Ubuntu 24.04 arm64 and x86_64: installed deb packages and extracted AppImage payloads passed the same workspace-screen, PI imports, SQLite, native PTY, authenticated daemon/Teams and MCP Wiki-search checks. The deb sandbox helper has root ownership and mode 4755.
Linux checks used containers and a virtual display; x86_64 ran under emulation on Apple Silicon. The x86_64 AppImage bootstrap could not execute in that environment, so its payload was extracted with unsquashfs for testing. Its bootstrap bytes match the previously released 0.7.0 runtime. These checks do not establish native x86_64 launcher or FUSE-mount behavior.
GitHub-hosted CI was unavailable; the verification above ran locally. Packaged smoke tests make no live model calls.
SHA-256 checksums
f4141503189b0b04754421ff426b499ea7ac91b8bcfd5eeb6cd3cd1b6fa40ce0 Research-Pilot-0.8.2-linux-amd64.deb
002118dd5b478a1e0aaa3cdf1a15badfaf21bac2f1f82dc3cab786a5ee994784 Research-Pilot-0.8.2-linux-arm64.AppImage
d235c30058d0222d9da773f5ca526be6af08ae1cb2470a21dcb619948f093696 Research-Pilot-0.8.2-linux-arm64.deb
f502b63c27cff2a655c09be4eea905f9f41743ec222f72d7da8ae3fe5002231d Research-Pilot-0.8.2-linux-x86_64.AppImage
bf9c55fcdf1251702decefce5949769ec95c3358a65f8d8ce96b7975ef1b0aad Research-Pilot-0.8.2-mac-arm64.dmg
f34dfbf5a8d892ba53d114768518f0d6300ed031e343863e4c90af5447e2298a Research-Pilot-0.8.2-mac-arm64.zip
Research Pilot 0.8.1
Research Pilot 0.8.1 improves Resident Researcher briefings and restores final manuscript acceptance after an accidental stop.
What changed
- Resident Researcher now prepares slide briefings covering work since the previous briefing, with background, work completed, progress, challenges and next steps. Writing guidance introduces unfamiliar terms and method designs before presenting comparisons. A separate slide-reading review helps identify missing context.
- Resident researchers have access to the shared research skills and core tools needed for literature, artifacts, experiments and presentations.
- The Resident desk now follows the Teams board layout, with clearer progress, reports and experiment views, plus motion that respects reduced-motion preferences.
- A stopped manuscript whose writing and review tasks are already complete can explicitly recover final acceptance through the management tool. Recovery retains the accepted package and prior records, starts a fresh coordinator conversation, and leaves the team paused until resumed. It does not automatically approve the manuscript or reopen unfinished writing.
- Agent launch setup now shares the workspace synchronization queue, preventing skill installation from racing workspace cleanup.
Installation and permissions
macOS Apple Silicon: open the DMG and drag Research Pilot into Applications.
Ubuntu/Debian: install the matching deb with sudo apt install ./Research-Pilot-0.8.1-linux-amd64.deb (use linux-arm64.deb on ARM). The deb is recommended. AppImage users on Ubuntu 24.04 may need libfuse2t64 and --no-sandbox; ARM AppImages also need zlib1g-dev. No Windows or Intel Mac installer is included.
Install and sign in to Claude Code and/or Codex CLI before using their tabs. Desktop and mobile launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents act with your user account's filesystem and network permissions. These flags do not grant administrator rights.
Quit the existing application before replacing it so background processes load the new code. Research data and settings remain under ~/.pipilot; back up important data before upgrading.
Verification
- Release gate: 1,737 tests passed, 7 skipped, zero failures; desktop type checking, release builds, size/dependency checks and the isolated MCP/daemon demo passed.
- macOS arm64: the signed app and DMG passed Apple notarization and Gatekeeper. Stapled tickets were verified, including the app extracted from the ZIP. Both the mounted DMG and extracted ZIP passed packaged renderer, PI imports, SQLite, native PTY, authenticated daemon/Teams database and MCP wiki-search checks.
- Ubuntu 24.04 arm64 and x86_64: installed deb packages and extracted AppImage payloads passed those same runtime checks. The deb sandbox helper has root ownership and mode 4755.
Linux checks used containers and a virtual display; x86_64 ran under emulation on Apple Silicon. They do not establish native x86_64 desktop or FUSE-mount behavior. No live model call was part of the packaged smoke tests.
SHA-256 checksums
e4b1b89f64f817f1b57fb9d6f9099fcdeb552bd14ddfac47b726f5ca2145351d Research-Pilot-0.8.1-linux-amd64.deb
c5433e4820b738b1675fe35268d7cc079afbd34b56bc39b82b414c092905d24b Research-Pilot-0.8.1-linux-arm64.AppImage
961c22b219565b0de3e2a68d0c0a2b000a6ee84d7348d554eccca6ff70bd8b41 Research-Pilot-0.8.1-linux-arm64.deb
c47b64e82585c47f4e63c975446437005c6392ff772c0bc81acc283c6220686a Research-Pilot-0.8.1-linux-x86_64.AppImage
a3cf7436951ab7bed1eafdd53ab38c7d385a39cd27219479777ed3214c0d7796 Research-Pilot-0.8.1-mac-arm64.dmg
6fb68f071947b1701a8c2b8592d61b7c8354b81531c09385e95701d3fe4c0524 Research-Pilot-0.8.1-mac-arm64.zip
Research Pilot 0.8.0
Research Pilot 0.8.0 adds collaborative Research and Manuscript Teams and improves recovery when an agent stops unexpectedly. Manuscript work now has one lead author, reusable support tasks, and a lighter acceptance review.
What changed
- Teams retain task state and messages across interruption and recovery. Usage is tracked without imposing a cumulative team budget; provider quotas and external charges still apply.
- A manuscript lead author owns the draft and delegates supporting work. Repeated checks can reuse existing tasks, and supporting agents have restricted write access to protect the manuscript.
- Final manuscript review no longer requires the coordinator to reread every file byte or every PDF page. Reading records can combine page ranges completed in any order.
- Wiki jobs have clearer progress and bounded full-text responses. PI sessions include recovery improvements for long conversations and provider changes.
- Release packaging now builds SQLite for each Linux target and keeps experiment-worker startup separate from the packaged daemon.
Installation and permissions
macOS Apple Silicon: open the DMG and drag Research Pilot into Applications.
Ubuntu/Debian: install the matching deb with sudo apt install ./Research-Pilot-0.8.0-linux-amd64.deb (use linux-arm64.deb on ARM). The deb is recommended. AppImage users on Ubuntu 24.04 may need libfuse2t64 and --no-sandbox; ARM AppImages also need zlib1g-dev. No Windows or Intel Mac installer is included.
Install and sign in to Claude Code and/or Codex CLI before using their tabs. Desktop and mobile launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents act with your user account's filesystem and network permissions. These flags do not grant administrator rights.
Quit the existing application before replacing it. Research data and settings remain under ~/.pipilot; back up important data before upgrading.
Verification
- Release gate: 1,715 tests passed, 16 skipped, zero failures; desktop type checking, release builds and the isolated MCP/daemon demo passed. Regression tests also start the minified daemon and resident worker as actual process entry points.
- macOS arm64: signed app and DMG accepted by Apple notarization and Gatekeeper. Stapled tickets verified in the DMG and extracted ZIP. Packaged renderer, PI imports, SQLite, native PTY, authenticated daemon/Teams database and MCP wiki search passed.
- Ubuntu 24.04 arm64 and x86_64: deb installation and extracted AppImage payloads passed the same runtime checks. The deb sandbox helper has the expected root ownership and 4755 mode.
Linux checks used containers and a virtual display, with x86_64 emulation on Apple Silicon. They do not establish native x86_64 desktop or FUSE behavior; the x86_64 AppImage filesystem was extracted directly for testing. No live paid model call or full real-paper Team run was part of this release smoke test.
SHA-256 checksums
b9ddaf0c87da2d6611a8436b6b79b9e14d9820b51c7808160f9262b2fef8cb01 Research-Pilot-0.8.0-linux-amd64.deb
7c76e910ce935adf6e5da7e8c3e4fdc1f40fbf900119cd20454bdf86ffad6ce1 Research-Pilot-0.8.0-linux-arm64.AppImage
97b72fad40ecc57d9981764abd0b862614a871fbe2d23c6592ad9876ea061244 Research-Pilot-0.8.0-linux-arm64.deb
5f8e4c5726948d3c3ea3c51f8bc68e2d4ca27964820dae9444a7ecab61861ac0 Research-Pilot-0.8.0-linux-x86_64.AppImage
67526bbe84bc3c720ccd57d89d924afd2bfb96c7431404fa0f2c5e75aa97a27f Research-Pilot-0.8.0-mac-arm64.dmg
d95c00397eb194907a60061a731870f4ef7852b84422d18909d85f900278f005 Research-Pilot-0.8.0-mac-arm64.zip
Research Pilot 0.7.0
Research Pilot 0.7.0 brings work modes, a plan-and-review workflow for research, and rebuilt context management for longer PI Agent sessions. It also updates the product documentation around how to use the application, with diagrams and a separate installation guide.
Work modes across agents
- Choose a workspace's default way of working across Claude Code, Codex and PI Agent. Collaborator is the general mode; experimental choices include Ideator, Builder, Teacher, and Policy & Records.
- Inspect a mode, copy it, or create a custom mode. Saving and applying are separate actions. Switching previews the foreground sessions it will end and starts fresh sessions with the new configuration.
- Mode editing now reports unsupported choices and failed switches accurately, protects shared modes from unsafe deletion, and asks before discarding unsaved edits.
- The phone companion exposes work mode under Settings. PI model and reasoning choices are remembered for the workspace after restarting the app.
Plan and follow research in one place
- Describe a research question on the Research page, generate a readable plan, request changes, and select Approve & start for the version you reviewed. Planning uses its own session and does not occupy ordinary PI chat.
- Draft inputs, revision requests and plan versions are saved. Generating or editing a plan does not start a run. A stopped or outdated planning response cannot replace the current plan.
- New runs can compare independent proposals, critique them, choose a direction, refine it and produce a readable design. They can also refine an existing direction directly. The coordinator retains responsibility for the next action.
- Design outputs and their dependencies are recorded, so replacing an upstream proposal marks affected downstream work for reconsideration. Earlier versions remain inspectable.
- Coordinators can retrieve detailed history and evidence as needed instead of receiving every record on every turn. The Research view presents the current narrative with supporting detail in a side panel.
These changes support recorded, reviewable investigations. They do not establish that a generated design is novel or that the design workflow improves scientific outcomes. The run's model-cost limit starts after launch; planning calls and external compute or service costs are separate.
Longer PI Agent sessions
- Context is checked at the model-request boundary, including tool definitions and the provider's reported usage.
- Large tool results can be stored by the local data service and replaced in working context with excerpts and retrieval references. Their original text remains available for follow-up reads.
- Compatible tool changes preserve compressed working context. Restoring a session and switching providers follow explicit compatibility rules.
- Codex subscription sessions reserve 32k tokens for local budget calculations and compact later. This does not add a response-length cap to the provider request. Anthropic sessions use a 300k working-window ceiling to bound context growth.
- The interface shows compression progress and elapsed time. Approval waits, queued follow-up instructions, file links and conversation history have received related fixes.
- Provider request handling includes fixes for cache-control block limits and diagnostic response-body handling.
Memory grounded in sources and time
- Memory records carry their originating user statement and relationships such as updates, historical facts and equivalence.
- A difference in wording alone no longer becomes a startup conflict warning. Disagreements are considered with their source, time and relevance to the current task.
- The online memory experiment has been removed. The full source-and-time relation behavior is the default when memory is enabled; work-mode settings and “restart without memory” remain available.
Documentation and release fixes
- The README now explains the product, first use, agent and mode choices, papers and memory, research planning, phone access, permissions and upgrades.
- Two diagrams show how project work becomes reusable knowledge and how a research plan reaches a reviewed result. Detailed setup and troubleshooting live in a separate guide.
- Settings loads when opened, reducing the startup JavaScript from about 484 KiB to 410 KiB and keeping it within the existing release budget.
- The Linux search-error test now accepts GNU grep's diagnostic wording and checks that malformed expressions remain visible errors.
Install or upgrade
Finish active work and quit Research Pilot before replacing the app. Research data remains separate under ~/.pipilot by default. Back up that directory and your project folders; do not remove them during an upgrade. Reopen the application so the new desktop and bundled data service are used together.
| Platform | Recommended | Alternative |
|---|---|---|
| macOS · Apple Silicon | DMG | ZIP |
| Ubuntu/Debian · Intel / AMD 64-bit | deb | AppImage |
| Ubuntu/Debian · ARM 64-bit | deb | AppImage |
Windows and Intel Mac installers are not included. Prefer the deb on Ubuntu. AppImages on Ubuntu 24.04 need libfuse2t64; arm64 also needs zlib1g-dev. Older Ubuntu releases use libfuse2. See the guide for the Chromium --no-sandbox fallback when AppImage startup is blocked by sandbox setup.
For Claude or Codex sessions, install the corresponding CLI and sign in. PI Agent's runtime is bundled and needs a configured model and authentication. Research Pilot itself needs no separate Node.js installation. Electron remains at 31.3.1.
Installation, configuration, phone setup and troubleshooting.
Agent permissions
Desktop and phone launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents can run commands, edit or delete files, and access the network with your OS account's permissions without per-action approval. Codex's own sandbox is disabled. This does not grant administrator privileges or bypass OS access controls. Use trusted workspaces and tasks, and keep backups.
Download a named application package above. GitHub's automatic Source code (zip) and Source code (tar.gz) entries contain public documentation, not the application source or an installer. The named mac-arm64.zip is the packaged Mac application.
Release validation
- Automated checks: 1,323 tests passed locally. GitHub CI passed the Node 18 test job and the Node 20 production release gate. The gate includes Electron typechecking, the production build, size and dependency budgets, and the daemon/MCP demo.
- Settings loading: the production renderer was exercised with isolated API fixtures. Settings was absent from startup requests, loaded on demand, displayed its Memory tab, and closed and reopened successfully.
- macOS packages: the application and DMG passed Developer ID signature, Apple notarization, stapled-ticket and Gatekeeper checks. Applications inside both the DMG and ZIP report 0.7.0 and pass signature and ticket validation. Their application archives match. The DMG application launched its actual bundled interface and opened Settings in an isolated profile.
- Linux packages: both debs installed in fresh Ubuntu 24.04 test containers and set the expected sandbox permissions. Both packaged interfaces displayed the workspace picker under a virtual display and started their bundled 0.7.0 data service.
- Bundled runtime: the Mac application, both installed Linux debs and both extracted AppImage payloads loaded the PI provider modules, spawned a native terminal, started the data service, rejected an unauthenticated API request, and completed a Wiki search through the packaged MCP adapter.
Linux graphical checks used a virtual display and Chromium's --no-sandbox flag inside containers. The x86_64 checks ran under emulation on Apple Silicon, not physical x86_64 hardware. The ARM AppImage extracted through its own launcher; the x86_64 payload was extracted directly because its launcher is not supported in this emulation setup. These packaging checks did not run a new authenticated model conversation or re-evaluate research-answer quality.
Research Pilot 0.6.10
Research Pilot 0.6.10 is a reliability release. Research runs no longer stall on experiments that can never report back, a delegated reviewer survives long quiet stretches, and a run's cost limit now holds even when several reviewers work at once. Terminals reattach with their colours and state intact, and phone attachments are stored privately.
Research runs
- A run no longer waits forever on an experiment whose status can never be refreshed, such as a backend that is no longer configured or has changed version. The coordinator is woken once with the reason, so it can re-run or report the job. When a backend is only temporarily unreachable, the run keeps waiting without spending model calls and tells you why.
- A delegated reviewer that is thinking for minutes inside one request is no longer mistaken for a stalled one. If a checkpoint was requested during such a quiet stretch, it is withdrawn once that reviewer resumes. Hung or looping sub-agents are still stopped by the overall time limit.
- The run's cost limit now holds with parallel reviewers. At most one reply's worth can go over the limit, however many requests run at once. Requests whose cost was never reported count at their maximum possible cost for the limit. A request that only has to wait for others to finish waits up to a minute instead of reporting the budget as spent.
- A run can no longer close as answered after an accepted code audit or design review behind its evidence was withdrawn.
Terminals
- Reloading the window, or opening a workspace in a second window, now restores each Claude and Codex tab faithfully. The screen, the most recent 3000 lines of history, colours, cursor, and terminal modes all come back. Stray fragments such as
36;231;221mno longer appear at the top of the history. - A terminal that hit an error while being shown now keeps following window resizes.
Phone and attachments
- Photos, files, and pasted images are now stored in
~/.pipilot/attachments, private to your account, instead of the shared temporary folder. Files older than 30 days are removed, and the folder is kept under 1 GB. - Two files with the same name no longer overwrite each other. Long names keep their extension, and long non-Latin names no longer fail to save on Linux.
- Uploads are written to disk as they arrive rather than held in memory. A file over 25 MB is refused on the phone before it is sent, and the phone now shows the size limit instead of a generic "upload failed".
Settings and the PI agent
- Saving Settings no longer resets the PI reasoning effort.
- Endpoints and API keys saved in Settings reach PI sessions in every window, including sessions started from the phone, without interrupting them. Web search picks up a newly saved Brave key immediately.
- The Default model list follows the self-hosted models editor as you edit it. If the saved default no longer exists, the list says what will happen instead of showing a different model.
- Self-hosted model entries edited by hand in
config.jsonthat cannot be used are kept as written and listed in Settings with the reason, instead of being deleted by the next save. - Form fields have a clearly visible edge in every theme, at a contrast of at least 3:1. Panel dividers stay quiet.
Memory
- A project memory saved again under its title is treated as current, instead of staying hidden because it had once been replaced.
Install or upgrade
Finish active tasks and quit Research Pilot before replacing the application. Research data remains separate from the application under ~/.pipilot by default.
| Platform | Recommended package | Alternative |
|---|---|---|
| macOS · Apple Silicon | mac-arm64.dmg | mac-arm64.zip |
| Ubuntu/Debian · Intel / AMD 64-bit | linux-amd64.deb | linux-x86_64.AppImage |
| Ubuntu/Debian · ARM 64-bit | linux-arm64.deb | linux-arm64.AppImage |
Windows and Intel Mac installers are not included. Prefer the deb on Ubuntu. AppImages on Ubuntu 24.04 require libfuse2t64; arm64 additionally requires zlib1g-dev. Older Ubuntu releases use libfuse2. If Chromium sandbox setup prevents AppImage startup, use the documented --no-sandbox fallback.
For Claude Code or Codex sessions, install the corresponding CLI and sign in once. No separate Node.js installation is required for Research Pilot itself. Electron remains at 31.3.1 in this release.
Detailed installation, permissions, and API setup guide.
Agent permissions
Desktop and phone launch Claude Code with --dangerously-skip-permissions and Codex with --dangerously-bypass-approvals-and-sandbox. Agents can run commands, modify or delete files, and access the network with your OS account's permissions without per-action approval. Codex's own sandbox is disabled. This does not grant administrator privileges or bypass OS access controls. Use trusted workspaces and tasks, and keep backups of important files.
Download one of the six named application packages above. GitHub's automatic Source code (zip) and Source code (tar.gz) entries contain the public README snapshot, not the application source or an installer. The named mac-arm64.zip asset is the packaged Mac application.
Release validation
For 0.6.10, 1062 automated tests passed on the shipped code, together with Electron typechecking, the production build, release size and dependency budgets, and the end-to-end demo. Every fix in this release has a regression test that fails on the previous code. The shipped bundles were checked to contain the new code.
The macOS application and DMG passed Developer ID signature, Apple notarization, stapled-ticket, and Gatekeeper checks. The applications inside both the DMG and the ZIP report version 0.6.10 and pass signature and stapled-ticket validation.
Both Linux debs were installed in clean Ubuntu 24.04 containers. Each installed the intended ALSA provider, set the Chromium sandbox permissions, shipped the native terminal library for its architecture, and started the packaged data service with appVersion 0.6.10. Both AppImage payloads passed the same architecture and service checks. The arm64 package was extracted with its own runtime. The x86-64 payload was extracted directly, because its AppImage runtime cannot execute under Apple Silicon Docker emulation.
Packaged graphical startup was not exercised on either platform. The terminal reattach, reviewer watchdog, and concurrent cost-limit changes were verified by tests against the real modules, not in a live app session. The x86-64 checks ran under emulation rather than on physical x86-64 hardware.