Pull Request Explorer

Summary

Parallelize semantic e2e validation

Speed up end‑to‑end tests by running semantic validation in parallel, reducing CI time and improving test reliability. This change enhances developer productivity by cutting test execution time.

Health Assessment

• • Fast turnaround with minimal review comments and quick merge indicates high confidence in the changes.

Summary

Refactor desktop binary manager for lazy install

Reduces installer size and ensures users run the latest binaries, improving reliability and user experience.

Health Assessment

• • The PR was merged within 3.2 hours with only a single review comment, indicating a smooth review process. The change is a large refactor affecting 20 files and over 3,200 lines, but the minimal review rounds suggest confidence in the implementation.

Summary

Add desktop‑specific User‑Agent for cloud requests

Enables LobeHub Desktop to identify itself to LobeHub Cloud, improving request routing and analytics.

Health Assessment

• • The PR was merged within 0.9 hours with no rework, indicating a smooth review process and low complexity.

Summary

Fix hetero execAgent user turn chaining

This fix ensures that follow-up execAgent turns in an existing topic are correctly chained to the current spine head, preventing orphaned root messages and improving conversation continuity.

Health Assessment

• • Quick fix with minimal changes and fast review, indicating low complexity and high confidence.

Summary

Fix GitHub MCP tool rendering

Adds dedicated GitHub inspector and render surfaces for Codex MCP tool calls, improving UI clarity and user experience.

Health Assessment

• • The PR was reviewed and merged within 24 minutes, showing minimal friction and a straightforward fix, despite the large code addition.

Summary

Remove obsolete subscription iframe wrapper

Clean up business setting pages by removing the unused SubscriptionIframeWrapper component and replacing affected pages with lightweight placeholders, improving maintainability and reducing bundle size.

Health Assessment

• • Merged within 0.6 hours with a single review comment, indicating a straightforward refactor with minimal impact on the codebase.

Summary

Fix task UI routing for running tasks

Corrects navigation and status display for running tasks, improving user experience and reducing confusion.

Health Assessment

• • PR addressed UI routing regressions with a single commit and no review rounds, indicating a streamlined process, but the large scope across frontend and backend suggests moderate risk for downstream impact.

Summary

Replace Lobe example ids with Linear ids

Update Linear entity renderer code to use standard Linear issue ID format, replacing Lobe example references.

Health Assessment

• • The PR was merged quickly with minimal review comments, indicating a straightforward and low-risk change.

Summary

Disable Claude Code wakeup tools

Disables problematic Claude Code tools across all spawn paths, preventing stuck wakeup behavior and ensuring consistent behavior in desktop and non-desktop environments.

Health Assessment

• • Fast cycle time of 2.2 hours, single review comment, minimal code changes, indicating low complexity and quick approval.

Summary

Refactor chat input for full-width menus

Improved user experience with full-width menus for @ mentions and / slash commands

Health Assessment

• • Fast cycle time and immediate review suggest a well-planned refactor
• • Large scope and removal of custom code may introduce some risk

Summary

Fix market skill discovery timeout

Adds a timeout to skill discovery to prevent slow or unavailable market connections from blocking agent execution, and guards client prefetch for unauthenticated users.

Health Assessment

• • Merged in under 20 minutes with no review rounds, indicating a straightforward bug fix with minimal risk.

Summary

Add market agent entry

Adds an "Add from Market" option to the agent creation menu, shares menu logic between the Agent section and header, and exposes a reusable model‑config endpoint path for consumers.

Health Assessment

• • Merged within 18 minutes of opening, with a single comment and no rework, indicating a smooth, low‑friction process.

Summary

Add constrained analytics query layer for CLI

Enables server‑side aggregation for message analytics, reducing client load and ensuring ownership‑scoped queries.

Health Assessment

• • Fast review and merge with minimal iterations; large code addition but well‑tested and ownership‑scoped.

Summary

Refactor: Replace LOBE markers with contextual comments

This refactor replaces legacy LOBE-XXX markers with inline comments that explain architectural decisions, improving code clarity. It reduces future maintenance overhead by making the rationale explicit.

Health Assessment

• • The PR was merged quickly with minimal review, indicating a straightforward refactor with clear intent and low risk.

Summary

Fix folding of completed agent turns

Ensures completed agent turns are folded correctly, improving UI clarity and performance for users.

Health Assessment

• • The PR was reviewed and merged within 1.6 hours with minimal back‑and‑forth, indicating a straightforward bug fix with low risk.

Summary

Fix: forward gateway approval mode to server runtime

Ensures gateway approval configuration is correctly passed to the server runtime, preventing misconfigurations and improving reliability.

Health Assessment

• • Rapid review and merge within 4 hours, minimal code changes, AI review contributed to quality.

Summary

Add autocomplete hints and keyboard links to forms

Enhances accessibility and mobile UX by enabling email keyboards, autofill, and keyboard navigation for form links, reducing friction for users.

Health Assessment

• • Fast cycle time of 1 hour with no rework; AI-generated code was reviewed quickly, indicating high confidence in quality.

Summary

Fix review panel group header shrink

Prevents UI layout issues in the review panel, ensuring stable header and row heights during expand/collapse, improving user experience.

Health Assessment

• • Rapid, single-commit fix with no tests added; minimal code churn and quick approval.

Summary

Fix Verify Runtime Isolation

Isolate verifier runtime from heterogeneous parents to prevent overwriting check results and improve Verify functionality.

Health Assessment

• • Fast review and merge, low number of commits and review rounds indicate a well-prepared and straightforward PR.

Summary

Add receipt revert support for agent signals

Enables users to revert agent signal receipts, improving error recovery and data consistency.

Health Assessment

• • Rapid merge with minimal review indicates high confidence but large code changes may increase risk.

Summary

Fix subagent thread title for reasoning-first events

Ensures subagent threads display the correct spawn description instead of the generic "Subagent", improving UI clarity and user experience.

Health Assessment

• • Quick AI-assisted fix with minimal rework; PR merged within 1.4 hours and required only a single review comment.

Summary

Fix agent builder tool conflicts

Ensures Agent Builder mode only uses its own tools, preventing self‑editing and improving reliability for users configuring agents.

Health Assessment

• • Quick, single‑commit fix with AI‑assisted review; minimal rework and fast merge.

Summary

Show model display name in message footer

Users now see human‑readable model names in chat footers, improving clarity and reducing confusion.

Health Assessment

• • Merged within minutes with a single comment, indicating a straightforward, low‑risk change.

Summary

Fix sub-agent elapsed time display

Corrects the elapsed time calculation for sub-agent tasks, ensuring timers accurately reflect the moment the tool was invoked and improving user experience.

Health Assessment

• • Fast turnaround with minimal changes and a single review comment, indicating low complexity and high confidence in the fix.

Summary

Enlarge updater modal actions for desktop

Adjusts button sizing in the desktop updater modal to improve usability while keeping notification buttons compact.

Health Assessment

• • Rapid review and minimal changes indicate low complexity and risk.

Summary

Add June 29 weekly changelog

Adds the weekly changelog entry covering the past 7 days, documenting new features, improvements, and fixes for the product.

Health Assessment

• • Quickly merged with minimal review, indicating a straightforward documentation update.

Summary

Release v2.2.9 with agent reliability and CLI updates

This release strengthens agent execution, adds CLI self‑update, and expands provider compatibility, delivering a more robust and feature‑rich platform for users and developers.

Health Assessment

• • Merged within 2 hours of opening, indicating a fast review cycle, but the PR introduces over 130k lines of code across 20 files, representing a high‑impact change that carries significant risk despite the short review period.

Summary

Fix chunk actions to use file IDs

Ensures chunk operations correctly reference underlying file IDs, improving consistency and preventing mismatches in knowledge resources.

Health Assessment

• • The PR received a review immediately but took over 3 days to merge, suggesting delays in final approvals or testing.

Summary

Standardize zh-CN translations for workspace

Ensures consistent terminology for workspace in Chinese locale, improving user clarity.

Health Assessment

• • Quick, single-commit PR with minimal changes, no code, low risk.

Summary

Expose taskTemplate router to mobileRouter

Allows the mobile client to call task template endpoints, enabling recommended task templates functionality.

Health Assessment

• • Fast cycle time and immediate review indicate a straightforward change with minimal risk.

Summary

Fix GLM-5.2 params and model descriptions

Corrects payload compatibility for Fireworks-hosted GLM‑5.2 and improves model‑switch UI by rendering descriptions, preventing API errors and enhancing user experience.

Health Assessment

• • Fast turnaround with minimal review comments indicates a straightforward bug fix; extensive localization updates across many files contribute to a large scope.

Summary

Add subtask topic activity to task detail

Enables users to view descendant task topics and source metadata in task detail, improving traceability and UI clarity.

Health Assessment

• • Rapid review and merge with minimal iterations, indicating a straightforward change that was well understood by the team.

Summary

Restore Update Notification UI

Reverts the update notification to its previous compact UI, improving user experience.

Health Assessment

• • The PR was reviewed and merged quickly, indicating a smooth development process.

Summary

Fix unavailable model notice in chat

Adds a warning when the selected chat model is no longer available, improving user experience by preventing confusion.

Health Assessment

• • Quick turnaround with minimal rework; AI reviewers provided automated feedback but no significant changes required.

Summary

Add global human-approval notification island

Enables users to approve or reject pending requests directly from a top‑center notification, eliminating the need to switch conversations and improving workflow efficiency.

Health Assessment

• • The PR was merged within 2 hours of opening, with a single review comment and no rework, indicating a smooth review process despite a large code addition.

Summary

Fix heterogeneous run verification flow

Ensures heterogeneous task runs correctly trigger verification, preventing silent failures and improving reliability.

Health Assessment

• • Fast review and merge with minimal iterations; quick cycle time indicates low complexity and high confidence in the change.

Summary

Fix agent skills refresh after skill import

Ensures newly imported skills are immediately visible to users by moving refresh logic to a lifecycle hook that runs on both client and server, improving user experience.

Health Assessment

• • Quick fix with minimal rework; automated review comments only; fast cycle time indicates smooth process.

Summary

Fix agent runtime verify for missing parent state

Prevents sub‑agent failures from leaving parent stuck, improving reliability.

Health Assessment

• • Quick, single‑commit fix with no rework; fast review and merge cycle.

Summary

Defer skill synthesis to turn completion

Enables accurate skill synthesis using full trajectory evidence, improving agent performance and reducing errors.

Health Assessment

• • Fast review and merge with comprehensive tests, but large code change indicates significant refactor.

Summary

Multiplex broadcast member streams onto supervisor WebSocket

Enables real‑time group chat by routing all member streams through a single supervisor WebSocket, improving performance and reducing client load.

Health Assessment

• • Fast cycle time and minimal review suggest high confidence in the implementation, though the large code change warrants careful monitoring for regressions.

Summary

Add deleteRun to remove published verification runs

This PR introduces a deleteRun mutation and CLI command to remove verification runs, ensuring data consistency. It prevents orphaned reports and cleans up related records automatically.

Health Assessment

• • Fast cycle with immediate merge after single review; minimal rework.

Summary

Fix disabling Zhipu tool stream option

Adds an option to disable the tool_stream parameter for Zhipu GLM channels, preventing errors when upstream services reject this field and improving compatibility.

Health Assessment

• • The PR was reviewed immediately and merged after a single small commit, indicating a straightforward fix with minimal risk.

Summary

Promote 'Don't Ask Again' to First-Class Option

Improves user experience by making the 'don't ask again' option prominent, reducing missed approvals and streamlining workflow.

Health Assessment

• • Fast review and merge with minimal changes; AI-assisted code generation streamlined the update.

Summary

Fix desktop update notification layout

Adjusts desktop update notification to fit sidebar, improving UI clarity and transition smoothness.

Health Assessment

• • Quick fix with no rework, resolved within 1.5 hours, minimal impact.

Summary

Fix final answer visibility in folded turns

Ensures the assistant’s final answer remains visible when a bookkeeping tool call follows a folded turn, improving user experience.

Health Assessment

• • Quick, single‑commit fix with no functional regressions; minimal code churn and fast review turnaround.

Summary

Fix server-runtime group broadcast rendering

Ensures group broadcasts render correctly as an in-bubble AgentCouncil, improving UI consistency and preventing orphaned messages.

Health Assessment

• • Fast cycle time and minimal review comments indicate smooth integration; the PR addressed a rendering bug with moderate code changes.

Summary

Fix mid-chain toolless prose step in AssistantGroup

Ensures continuous assistant run rendering by merging toolless prose steps, improving user experience and consistency.

Health Assessment

• • Quick fix with minimal review, merged within 1.2 hours, indicating low complexity and high confidence.

Summary

Localize task run controls and show step count

Adds missing i18n keys for the stop run modal and updates the turn process fold to display step count, improving Chinese UI consistency and user clarity. These changes enhance user experience by ensuring all UI text is localized and informative.

Health Assessment

• • Fast cycle time and immediate review indicate a smooth process; AI assistance helped expedite localization and UI consistency improvements.

Summary

Fix concurrent append by anchoring new user turn

Ensures new user messages correctly reference the latest assistant turn, preventing conversation splits and improving consistency.

Health Assessment

• • Quick fix with minimal review; automated AI review helped catch issues and confirm correctness.

Summary

Fix gateway init for local hetero reconnect

Ensures streaming continues when reconnecting to local heterogeneous agents, improving reliability for users.

Health Assessment

• • Fast review and merge with minimal rework, indicating a straightforward bug fix that resolved a critical streaming issue.