Pull Request Explorer

Summary

Fix branch checkout routing into worktree

Improves user experience by routing branch checkouts to the correct worktree instead of failing, eliminating error toasts and simplifying navigation.

Health Assessment

• • Fast review and merge with minimal changes; single commit resolved issue quickly.

Summary

Health Assessment

Summary

Add new AI models to model bank

Adds support for newly released GPT‑5.6, Grok 4.5, Sonnet 5, and Nano Banana 2 Lite models, updating pricing tiers and runtime parameters, improving model card accuracy and fixing a pricing bug.

Health Assessment

• • The PR was reviewed and merged within 1.4 hours, indicating a smooth process with minimal friction. AI‑powered review comments were quickly addressed, resulting in rapid iteration and a straightforward merge.

Summary

Bump @lobehub/cli dependency

Updates the CLI tool to the latest patch, ensuring compatibility and bug fixes for users.

Health Assessment

• • Quick dependency bump with minimal changes, no review friction.

Summary

Add downloads hub for apps, CLI, Messenger

Provides a unified downloads hub, improving user experience by centralizing app, CLI, and Messenger acquisition.

Health Assessment

• • The PR added extensive localization files with minimal code changes, reviewed quickly by AI tools, and merged within a day.

Summary

Remove local sleep before QStash publish

Improves queue performance by eliminating in‑process delay, delegating sub‑second scheduling to QStash, reducing server latency.

Health Assessment

• • Merged within 0.6 hours with no review, indicating a straightforward performance tweak.

Summary

Fix execAgent anchoring to prevent orphan roots

Ensures new user turns correctly anchor to existing conversation, eliminating reversed message order and reducing orphan root creation.

Health Assessment

• • Fast 3.2‑hour cycle with a single commit and no review friction; extensive test coverage mitigates risk.

Summary

Fix unread dot visibility for topics with linked PR

Ensures users see unread indicators even when topics have linked pull requests, improving notification clarity.

Health Assessment

• • Quick turnaround with minimal review, indicating straightforward bug fixes and documentation updates.

Summary

Graduate floating chat panel out of labs

The floating chat panel in agent document preview is now always on, removing the labs flag and simplifying settings for users.

Health Assessment

• • Fast cycle time (2.5h) with minimal rework; AI assistance helped streamline the change; the PR impacted 20 files and 246 lines, indicating a moderate scope but low risk due to clear testing and no breaking changes.

Summary

Add conversation analytics business hook

Introduces a generic conversation analytics hook and user activity summary, providing business insights without altering core conversation behavior.

Health Assessment

• • Fast review and merge (6 h cycle) with minimal rework; the PR added a small number of lines across both backend and frontend, indicating low risk and high efficiency.

Summary

Fix outdated docs across site

Updated documentation to reflect recent refactors, ensuring accurate guidance for users.

Health Assessment

• • Single commit, rapid review, docs-only change with no code impact, minimal risk to production

Summary

Tighten worktree icon to branch name gap

Improved visual spacing in chat input's git status row, reducing the gap between the worktree icon and branch name pill for a cleaner UI.

Health Assessment

• • Quick single-commit PR with minimal changes, merged within 1.7 hours, indicating low complexity and minimal review friction.

Summary

Fix duplicate git branch icon in chat input

This change removes visual duplication of the git branch icon in the chat input control bar, improving clarity for users working with non‑GitHub repositories.

Health Assessment

• • Quick fix with minimal changes, merged within 1.1 hours, indicating low complexity and high confidence.

Summary

Fix non-ASCII handling in base64 helpers

Ensures base64 encoding/decoding works correctly in browsers for non-ASCII text, preventing crashes and data corruption.

Health Assessment

• • Quick review and merge within 2 days, minimal changes, low risk.

Summary

Fix surrogate pair handling in sanitizeUTF8

Corrects sanitizeUTF8 to preserve valid emoji and astral-plane characters, preventing loss of content in RAG document indexing.

Health Assessment

• • Quick fix with minimal changes, no rework, merged within 2 days.

Summary

Split Skills and Connectors into separate settings pages

Separates skills and connectors into dedicated pages, improving clarity and navigation for users.

Health Assessment

• • Fast review and merge with minimal iterations indicates a smooth process. Single commit and quick approval suggest low complexity.

Summary

refactor: extract server call llm stream sink

Separates stream accumulation from executor logic, enabling clearer boundaries and easier future enhancements, which reduces maintenance overhead and potential bugs.

Health Assessment

• • Rapid iteration with minimal review indicates high confidence in the refactor and low risk of regressions.

Summary

Refactor Agent for Configurable Graph Runtime

Redesigns the agent runtime to support declarative graph execution, improving modularity, tool integration, and developer flexibility.

Health Assessment

• • Rapid AI‑assisted review enabled quick fixes, but the large code churn and extensive test suite indicate a significant architectural shift with elevated risk.

Summary

Fix worktree switcher logic for main worktree

Corrects worktree switcher behavior, ensuring accurate icon display and delete options, improving user experience and preventing errors.

Health Assessment

• • Fast turnaround with minimal rework; quick review and merge within 3 hours.

Summary

Style: lead PR number in topic hover card

Improves visibility of PR numbers in the topic sidebar hover card by moving the number before the title and adjusting hover styling, ensuring stable identifiers are not truncated.

Health Assessment

• • Quick AI-generated style fix with minimal scope and fast cycle time.

Summary

Fix CLI model display during streaming

Ensures users see the model name and token usage in real‑time, improving transparency and debugging during agent conversations.

Health Assessment

• • Quick fix with minimal rework, fast review, and small scope indicates low risk.

Summary

Add worktree inspectors and state recording

Enables Claude Code to correctly display worktree actions in chat and maintain accurate topic state, improving user experience and reducing confusion.

Health Assessment

• • PR completed in under 10 hours with no major review back‑and‑forth, indicating a straightforward implementation.

Summary

Upload image results for CC Read on desktop

Enables Claude Code’s Read tool to upload and display image thumbnails on the desktop, improving user experience and reducing data transfer by replacing raw placeholders with rendered images.

Health Assessment

• • Merged within 1 hour of opening, indicating a smooth review process; extensive test coverage mitigates risk despite the large code changes.

Summary

Fix agent support for GPT-5.6 Codex models

Adds GPT‑5.6 Codex model variants, improves usage tracking, and restores session baselines, enhancing user control and accurate token accounting.

Health Assessment

• • Rapid review and merge with minimal iterations indicates high confidence in changes.

Summary

Persist activeScopeKey for parallel cache hydration

Improves boot performance by persisting cache scope, reducing load times and eliminating anonymous orphaning.

Health Assessment

• • Fast review and merge within 10 hours, minimal rework, significant performance improvement.

Summary

Fix gateway notification body to show streamed text

Ensures desktop notifications display the actual streamed report instead of a generic message, improving user experience.

Health Assessment

• • Quick fix with minimal changes, reviewed immediately, no rework needed.

Summary

Fix heterogeneous device exec argument forwarding

Ensures correct model and effort parameters are passed to remote device execution, improving reliability of heterogeneous Claude Code runs.

Health Assessment

• • Quick fix with minimal changes and immediate approval, indicating low complexity and high confidence.

Summary

Add verify check overview summary

Adds a panel to the verify report viewer that summarizes passed, failed, and uncertain checks and provides next‑step guidance, improving user clarity and workflow efficiency.

Health Assessment

• • The PR was merged within 9.6 hours with a single commit and no back‑and‑forth reviews, indicating a smooth, low‑friction process. The AI‑generated code was reviewed quickly and accepted, and the change touches only a few frontend files, keeping risk minimal.

Summary

Hide model shortcuts on home page

Removes deprecated model shortcut list from home page, reducing UI clutter while keeping code for future cleanup.

Health Assessment

• • Quick style change merged within 24 minutes, minimal code churn, no review required.

Summary

Fix detection of Codex CLI in ChatGPT.app

Local Codex agents can now launch from either the current ChatGPT.app bundle or the legacy Codex.app bundle, improving reliability for users.

Health Assessment

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

Summary

Bump CLI dependency to 0.0.39

Updates CLI dependency version, ensuring compatibility with upcoming release, no functional changes.

Health Assessment

• • Quick dependency bump with minimal changes, merged within 8 hours.

Summary

Add GPT-5.6 support with reasoning modes

Adds GPT‑5.6 model support, including reasoning modes and usage handling, improving AI integration for users.

Health Assessment

• • Rapid merge with minimal review, extensive test coverage, and clear documentation indicates a smooth process.

Summary

Refactor server call LLM context build

Extract server-side call LLM context hint resolution and context-engine input assembly to improve code organization and prepare for future splits.

Health Assessment

• • The PR was merged quickly, but its large scope and the need for follow-up PRs may indicate potential risks or complexities.

Summary

Refactor call_llm tooling resolution

Extracts call_llm step-level tool and skill resolution into serverCallLlmTooling, improving code organization and maintainability.

Health Assessment

• • The PR was merged quickly, indicating a smooth review process.
• • The size of the PR is relatively small, with only 3 files changed.

Summary

Fix serialization of CC ask-user interventions

Ensures user interventions are reliably queued and routed, preventing UI stalling and ensuring correct tool execution.

Health Assessment

• • Rapid resolution with minimal review, indicating a straightforward bug fix that required only a few iterations.

Summary

Add Claude SDK streaming lab

Adds a streaming SDK for Claude Code, enabling background task notifications and real‑time status monitoring for users.

Health Assessment

• • Fast cycle time and minimal review iterations indicate smooth integration and low friction.

Summary

refactor: move call_llm shell into runtime package

Refactors the call_llm shell into the runtime package, streamlining LLM call context handling and improving maintainability for future feature development.

Health Assessment

• • Merged within 16 minutes of opening, the PR required only a single comment and no additional commits, indicating a low‑friction, straightforward refactor with minimal risk.

Summary

Fix frontend state for hetero tool reconciliation

Ensures consistent frontend state during heterogeneous tool execution, improving reliability and reducing stale data races. This change prevents UI inconsistencies and race conditions during parallel tool batches.

Health Assessment

• • Rapid review and merge with minimal iterations indicates low complexity and high confidence.

Summary

Refactor call_llm routing via runtime transport

Improves modularity and testability of LLM call logic, enabling future feature expansion without server coupling.

Health Assessment

• • Fast turnaround with a single review comment and minimal rework; the PR was merged within 0.4 hours of opening.

Summary

Fix task scheduling stalling due to errors

Ensures scheduled tasks resume after manual failures and transient errors, improving reliability and auditability.

Health Assessment

• • Large change with slow cycle, but quick review; likely complex logic changes.

Summary

Migrate compress context executor to agent-runtime

Migrate compress_context into @lobechat/agent-runtime as a package-hosted executor, enhancing compression execution and server adapters.

Health Assessment

• • The PR was merged quickly with minimal review time, indicating a well-structured and clear change.

Summary

Health Assessment

Summary

Refactor: Remove Group Orchestration Task Messages

Simplifies group orchestration by eliminating placeholder messages, improving clarity and reducing noise for users.

Health Assessment

• • Fast cycle time with minimal review iterations indicates efficient collaboration.

Summary

Add full action URLs and wider modal

Enhances notification inbox by enabling external URLs to open in new tabs and expanding modal width for better readability.

Health Assessment

• • Quick review and merge with minimal changes, indicating straightforward UI polish.

Summary

Fix workspace scope for LH agent edits

Ensures LH agent edits correctly maintain workspace context, preventing misrouted commands and false success responses.

Health Assessment

• • Fast cycle time and minimal review comments indicate a straightforward bug fix with comprehensive tests, resulting in low risk to the codebase.

Summary

Fix optional builder context access in test

Ensures type safety in agent selector tests, preventing TS errors during build and maintaining test reliability.

Health Assessment

• • Quick fix with minimal changes and immediate approval, indicating low complexity and high confidence.

Summary

Fix lint: migrate restricted UI imports

This PR resolves lint failures by updating UI component imports and adjusting tests, ensuring compatibility with the latest UI package restrictions.

Health Assessment

• • Quick review and merge within 1.3 hours, minimal back‑and‑forth, indicating a straightforward lint fix.

Summary

Add three-state plugin config

Introduces a third state to per-agent plugin config, allowing for more flexible plugin management and improved user experience.

Health Assessment

• • The PR has a large scope with many files changed, but the review process was relatively smooth with no major issues raised.

Summary

Refactor Composio runtime to use connector metadata

Updates Composio runtime to source truth from connector metadata, eliminating reliance on the plugin table and improving reliability for connected accounts.

Health Assessment

• • Fast cycle time and minimal review comments indicate efficient collaboration, despite the large code change. The refactor improves runtime reliability without introducing new features or breaking existing behavior.

Summary

Refactor sub-agent executors into agent-runtime

Consolidates sub‑agent execution logic into the agent‑runtime library, simplifying server architecture and improving code reuse.

Health Assessment

• • Merged in under two hours with a single review comment, indicating a straightforward refactor with minimal complexity.