Pull Request Explorer

Summary

Enable disabling of builtin tools

Provides users with the ability to disable built‑in runtime tools and skills, enhancing customization and control over the chat experience.

Health Assessment

• • Initial review was rapid, but the PR required several rounds of comments and fixes, indicating moderate complexity and a need for careful testing of the new disabling logic.

Summary

Fix preservation of thinking content semantics

Restores clear separation between raw thinking text and final reasoning, preventing data contract drift and ensuring accurate model outputs. This improves reliability of AI responses and maintains compatibility with existing workflows.

Health Assessment

• • Merged within 2 hours with no review, indicating a straightforward bug fix with minimal risk and low friction.

Summary

Revert context size display change

Reverts a recent change that incorrectly displayed context size, restoring accurate token usage metrics for users.

Health Assessment

• • Revert merged within minutes with minimal review, indicating a straightforward change with low complexity and high confidence.

Summary

Health Assessment

Summary

Redesign desktop tray menu for consistent UX

This PR unifies tray interactions and adds dynamic navigation, giving users quick access to recent agents and improving workflow.

Health Assessment

• • Fast 3.5‑hour cycle with no review delays, but the change touches 20 files and 645 lines, indicating a significant UI overhaul. The update is low risk as it only affects desktop UI behavior.

Summary

Document webview screenshot fallback behavior

Clarifies screenshot fallback logic to reduce debugging effort and improve reliability of host-page CDP screenshots.

Health Assessment

• • Quick documentation update with no code changes, minimal review effort, and negligible risk to production.

Summary

Health Assessment

Summary

Open verify reports in portal

Allows users to view verification reports directly within the portal, improving transparency and streamlining workflow.

Health Assessment

• • The PR was merged in just over six hours with an AI‑driven review, but required several commits after the first review, indicating moderate rework. The change spans 19 files and 272 lines, representing a sizable feature addition.

Summary

Add local Claude Code / Codex session import backend

Enables one-click import of local CLI agent sessions into LobeHub topics, improving data ingestion and user workflow.

Health Assessment

• • Rapid review and merge (1.9h cycle time) indicates high confidence; minimal rework; no blockers.

Summary

Render Read image preview at natural aspect ratio

Fixes distorted image previews in the Claude Code Read tool by passing maxHeight directly to the image element, ensuring natural aspect ratio and preventing cropping.

Health Assessment

• • The PR was merged within 2 hours, involved a single line change, and required no additional tests, indicating a straightforward, low‑risk fix.

Summary

Upgrade @lobehub/ui dependency to 5.20.2

Updates the UI component library to the latest version, ensuring compatibility and applying updated lint rules for a smoother development experience.

Health Assessment

• • Rapid merge with minimal changes and no review indicates a routine dependency update.

Summary

Fix UI: Clarify Working Directory and Header

Improves user experience by correcting labels and removing ambiguous save indicators, ensuring clearer navigation and state management.

Health Assessment

• • Fast cycle time and minimal rework indicate a straightforward UI fix with low risk to production.

Summary

Fix rate-limited hetero run resume behavior

Ensures that when a Claude or Codex run hits a usage limit, the system resumes from the last successful step instead of deleting the entire conversation, preserving user progress and reducing duplicate work.

Health Assessment

• • The PR was reviewed and merged within 34 hours, with a single commit and no back‑and‑forth, indicating a smooth, low‑friction process. The change involves a large number of lines but is confined to frontend logic, mitigating risk.

Summary

Fix chat context size display

Corrects token usage display in multi-step agent runs, ensuring accurate context size and billing metrics.

Health Assessment

• • Quick fix with minimal changes, merged within minutes, indicating low complexity and high confidence.

Summary

Enrich Claude Code terminal error reporting

Improves error visibility for Claude Code agents, giving users clearer diagnostics and actionable details.

Health Assessment

• • Fast turnaround with a single commit and no review, indicating a straightforward, low-risk change.

Summary

Enable readFile to return images for vision models

Allows the local-system readFile tool to upload image files and provide image URLs, enabling vision models to process images directly and improving AI perception for desktop users.

Health Assessment

• • Fast review and merge with minimal back‑and‑forth; large code change but well‑tested and AI‑assisted review helped catch issues early.

Summary

Add in-app browser tab to WorkingSidebar

Adds an embedded Chromium browser as a new tab in the agent chat right-side WorkingSidebar, enabling users to browse web content within the desktop app without leaving the chat interface.

Health Assessment

• • The PR was reviewed and merged within 10 hours, indicating efficient collaboration and minimal blockers despite the large code additions.

Summary

Expose agent documents to mobile

Enables the mobile client to retrieve agent documents via the existing tRPC router, extending mobile functionality without altering web data paths.

Health Assessment

• • Rapid 0.9‑hour cycle time and only two lines added suggest a low‑risk, straightforward change.

Summary

Prevent cross-agent profile writes

Ensures edits remain scoped to the correct agent, preventing data corruption and improving user experience.

Health Assessment

• • Fast turnaround with minimal review, but large code changes increase risk of regressions.

Summary

Fix workspace device settings navigation

Ensures navigation to workspace device settings works correctly, improving user experience on desktop.

Health Assessment

• • Quick review and merge, minimal changes, low risk.

Summary

Refactor workspace_user_settings PK to surrogate id

This change replaces the composite primary key with a surrogate UUID primary key, improving scalability and simplifying future schema changes.

Health Assessment

• • Fast cycle time and minimal review indicate smooth integration; large line addition is due to migration metadata but no major blockers were encountered.

Summary

Fix runtime unresolved tool calls

Restores execution of fully-qualified tool calls, preventing empty assistant responses and improving reliability of the agent runtime.

Health Assessment

• • Fast turnaround with minimal changes, single review, no major rework, low risk to production.

Summary

Add localized bot connection error handling

Enables users to see localized, actionable bot connection errors, improving UX across languages.

Health Assessment

• • Fast review and merge within 3 hours, minimal rework, indicating a smooth process.

Summary

Fill missing locale translations

Adds missing translations for multiple locales, ensuring consistent UI language support.

Health Assessment

• • Rapid review and merge indicate a straightforward data update with minimal risk.

Summary

Add SuperGrok provider via xAI OAuth

Adds a new SuperGrok provider enabling users to access Grok models through xAI OAuth device flow, eliminating the need for API keys and supporting token refresh.

Health Assessment

• • Fast review and merge with minimal back‑and‑forth, but the PR introduces a large number of lines across many files, indicating a substantial feature addition.

Summary

Add workspace_user_settings table and device visibility/share columns

Adds new database table and columns to support workspace‑scoped user settings and device visibility, enabling fine‑grained control over device sharing and visibility across workspaces.

Health Assessment

• • Fast cycle time of 1.8 hours and minimal review comments indicate a smooth integration.

Summary

Fix Codex login-shell wrapper for git detection

Restores accurate git command parsing in Codex, ensuring branch and PR tracking work correctly for users.

Health Assessment

• • Quick, single-commit fix with minimal rework; no significant review or iteration required.

Summary

Add local S3 storage for agent testing

Provides a local S3 server for agent tests, enabling realistic storage interactions and CORS validation.

Health Assessment

• • Fast 1.3‑hour cycle, minimal rework, straightforward addition of test tooling.

Summary

Improve message refresh cache consistency

Ensures consistent conversation state, reduces duplicate requests, and improves user experience during message refreshes.

Health Assessment

• • Rapid iteration with multiple commits after AI review indicates a tight feedback loop, but the large code churn and many files suggest a high-risk change.

Summary

Refactor: Extract LLM Attempt Transport

Improves modularity and testability of LLM call logic, enabling better error handling and blob storage integration.

Health Assessment

• • Merged within 1.7 hours with a single review comment, indicating a straightforward refactor with clear tests and minimal friction.

Summary

Refine rate limit recovery guidance

Improves user guidance for agent rate limits by updating UI states and localizing messages, enhancing clarity and usability.

Health Assessment

• • Fast review cycle with AI assistance, minimal rework, and clear scope indicate a low-risk, high-efficiency PR

Summary

Fix heterogeneous agent usage surface

Corrects live usage reporting for heterogeneous agents, improving accuracy and user visibility.

Health Assessment

• • Fast 1.7‑hour cycle, no review iterations, minimal rework, straightforward bug fix.

Summary

Add internal entity link previews

Enables preview of internal entity links, improving navigation and context for users within the chat interface.

Health Assessment

• • Rapid review and merge with minimal iterations indicate high confidence and low risk.

Summary

Remove verify report overview sidebar

Simplifies report layout by removing the sidebar, improving content width and reducing clutter.

Health Assessment

• • Single commit, no review, quick merge indicates low complexity and minimal risk.

Summary

Fix Claude Code retry loading state

Adds UI feedback for retrying API calls and disables flickering animation for tool messages, improving user experience during backoff.

Health Assessment

• • PR addressed UI and streaming issues with quick review turnaround, moderate code changes, and minimal friction.

Summary

Add Claude Code SDK runtime toggle

Enables users to switch the Claude Code SDK runtime on desktop via a Labs setting, giving them control over session behavior without relying on environment variables.

Health Assessment

• • The PR was merged within 1.6 hours with a single review comment, indicating a smooth, low‑friction process and minimal risk.

Summary

Refactor LLM call boundary for clarity

Improves LLM call orchestration, reducing complexity and enabling future decomposition.

Health Assessment

• • PR completed in under 4 hours with no review comments, indicating a smooth process.

Summary

Fix zero-row update reporting as success

Ensures database updates correctly flag failures, preventing orphaned message chains and data loss.

Health Assessment

• • Fast review and merge, minimal rework, quick resolution of a critical data consistency bug.

Summary

Fix worktree recording of literal $WT

Prevents incorrect worktree entries from shell variables, improving reliability of worktree switching.

Health Assessment

• • Quick fix with minimal rework, fast review, low risk.

Summary

Unify task/topic execution status icons

Consolidates status icons across UI surfaces, improving consistency and reducing duplication. This refactor centralizes icon logic, simplifying maintenance.

Health Assessment

• • Merged quickly with no review, indicating low complexity and high confidence in the change.

Summary

Upgrade Zod to v4

Upgrades the Zod validation library to version 4, ensuring compatibility across the codebase and updating related schemas and tests, improving type safety and future-proofing the project.

Health Assessment

• • Merged within 16 hours with a single automated review, indicating a smooth process and minimal friction.

Summary

Fix desktop bundle serving for /verify on mobile

Ensures verification pages load the desktop SPA on mobile devices, providing a consistent user experience and preventing mobile‑specific routing errors.

Health Assessment

• • Quick, single‑file fix merged within 1.6 hours with no rework or comments.

Summary

Fix analytics hook identity stability

Ensures stable hook identity to avoid redundant updates, improving performance and reliability of analytics tracking.

Health Assessment

• • Quick, low-impact bug fix with minimal changes and no review friction.

Summary

Refactor: Remove Legacy Server Call LLM Setup

By eliminating redundant message preparation logic, the codebase becomes leaner and more reliable. The clearer separation of concerns reduces maintenance overhead and potential drift between server and client behavior.

Health Assessment

• • Quick refactor with minimal changes, no blockers, merged within 30 minutes.

Summary

Refactor: extract server call LLM attempt

Improves reliability and maintainability of LLM call logic by isolating a single model call attempt, reducing code size and enabling better error handling.

Health Assessment

• • The PR was completed in under 11 hours with minimal review friction, indicating a well-scoped refactor that streamlined LLM call logic and reduced executor complexity.

Summary

Fix redundant report revalidation in verify

This change eliminates redundant background validation for verification reports, reducing unnecessary network traffic. It improves performance and ensures users see up-to-date data without flickering.

Health Assessment

• • The PR was merged in 0.4 hours with no review comments, indicating a straightforward fix with minimal friction.

Summary

Refactor: Extract Server Call LLM Finalizer

Isolates LLM call finalization logic to improve modularity, testability, and persistence handling, reducing executor complexity and enabling clearer state management.

Health Assessment

• • Fast 6‑hour cycle and a single AI‑review comment indicate a smooth refactor with minimal friction.

Summary

Fix OpenAI context limit preflight errors

Preflight checks prevent oversized prompts from causing provider errors, improving reliability and user experience.

Health Assessment

• • The PR addressed context limit errors with minimal changes and quick review, indicating a straightforward bug fix.

Summary

Release 20260710

This release introduces new agent capabilities, improved verification, and performance enhancements across CLI, desktop, and chat, delivering a more reliable and feature‑rich LobeHub experience.

Health Assessment

• • Rapid cycle time indicates efficient review, but the massive code changes suggest significant impact and potential for downstream issues.

Summary

Add user interaction and Codex error inspectors

Enhances user experience by providing interactive prompts and clearer error feedback for Codex integration.

Health Assessment

• • PR closed in 3.4 hours with a single review comment, indicating a smooth, low‑friction process and minimal rework.