Pull Request Explorer

Summary

Fix gateway buffer for high-frequency updates

Improves performance by buffering high-frequency gateway updates, reducing UI thread load and ensuring smoother user experience.

Health Assessment

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

Summary

Add CLI trace inspect and replay

Adds CLI commands to inspect and replay recorded LLM operations, enabling debugging and model comparison.

Health Assessment

• • The PR was merged within 12.4 hours after a single review, but required multiple commits to address feedback, indicating a fast but iterative process.
• • AI review was used to provide automated suggestions, which helped accelerate the review cycle.

Summary

Expand LobeHub provider support for heterogeneous agents

Adds support for multiple LobeHub providers (Claude, Codex, Grok, Kimi, Pi) enabling local agents to use deployment‑provided models, improving flexibility and reducing configuration overhead.

Health Assessment

• • Rapid review and single iteration indicate confidence in changes; large code addition suggests significant feature addition.

Summary

Sign agent trace snapshots for trace CLI

Adds a read‑only agent trace API that signs S3 URLs and lists operations, enabling secure, scoped access to execution snapshots for the trace CLI.

Health Assessment

• • Fast cycle time and minimal review comments indicate a straightforward, low‑risk implementation.

Summary

Refactor: Remove Obsolete Subscription Mismatch Error

Eliminates an unreachable error condition, simplifying error handling and reducing maintenance overhead.

Health Assessment

• • Quick refactor with minimal changes, merged within minutes, indicating low complexity and risk.

Summary

fixed local-file preview toolbar

Replaces floating overlay controls with a fixed toolbar, improving usability and responsiveness for local file previews.

Health Assessment

• • The PR was reviewed quickly with AI assistance, but required multiple commits and review rounds, indicating moderate complexity.
• • The new toolbar component enhances user experience for local file previews, adding a visible path and responsive action buttons.

Summary

Record actor for Goal Graph transitions

Adds accurate attribution of who performed each transition in goal graphs, separating user actions from system decisions, improving traceability and analytics.

Health Assessment

• • Fast 2.5‑hour cycle with a single review and no rework, indicating smooth integration of a moderate‑size change.

Summary

Fix image tool result reporting in context engine

Ensures successful image-generating tool calls are correctly forwarded to vision-capable models, preventing false failure reports and improving agent reliability.

Health Assessment

• • Quick, targeted fix with minimal code changes; resolved regression and restored correct multimodal handling.

Summary

Clean up UI mocks and improve test performance

This PR removes redundant per-file UI mocks, centralizes mock setup, and prebundles the UI library, cutting test runtime by 35% and simplifying maintenance.

Health Assessment

• • Fast review and quick iteration; AI assistance helped streamline mock cleanup and performance improvements.

Summary

Support direct site uploads

Adds ability to upload entire sites directly from the portal, improving content publishing workflow.

Health Assessment

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

Summary

Fix artwork reference image normalization

Ensures artwork generation uses correct image sources, improving visual quality and user experience.

Health Assessment

• • PR required multiple commits to address image source handling, indicating moderate complexity but resolved within a day.

Summary

Release 20260827: Agent Enhancements and Desktop Updates

This release introduces provider‑backed heterogeneous agents, restart‑free desktop updates, agent self‑learning, and improved acceptance workflows, enhancing user productivity and reliability.

Health Assessment

• • Rapid review and merge after 30 hours, indicating an efficient process, but the large code churn suggests a significant feature rollout with moderate risk.

Summary

Add process control UI and server-driven execution

Enables autonomous goal execution, reducing manual ticks and improving reliability for long-horizon tasks.

Health Assessment

• • Fast 19‑hour cycle with active review and multiple iterations, indicating high developer engagement but also significant rework; overall risk remains moderate due to rapid resolution and comprehensive testing.

Summary

Fix chat adapter fetchData type mismatch

Ensures adapters correctly handle the new chat library’s fetchData returning Buffer or ArrayBuffer, preventing CI failures and maintaining compatibility.

Health Assessment

• • Quick AI-assisted fix with minimal changes, no review required, merged within 30 minutes.

Summary

Fix subtopic creation on gateway send path

Ensures new subtopics are correctly created and displayed, improving thread navigation and user experience.

Health Assessment

• • The PR addressed a critical thread creation bug with extensive changes across backend, frontend, and tests, but was reviewed quickly and merged within a day, indicating efficient collaboration.

Summary

Fix chat input tag style overrides

Ensures upload status chips display correctly after UI component migration, improving user experience.

Health Assessment

• • Quick, minimal change that resolved a styling issue promptly.

Summary

Pin seeded message timestamps in execAgent

This change pins seeded message timestamps to avoid flaky failures in execAgent topic reuse. It improves test reliability and ensures consistent behavior in production.

Health Assessment

• • Fast turnaround with minimal changes, no production code touched, low risk to system.

Summary

Fix media refs for tool result images

Adds stable opaque media references for images produced by tool results, improving security and reliability by avoiding inline base64 and ensuring proper handling across contexts.

Health Assessment

• • Rapid iteration with AI‑assisted review led to quick resolution; minimal friction and low risk.

Summary

Fix chat image download via file proxy

Ensures chat image previews can be downloaded without CORS errors by using the file proxy download option, improving user experience.

Health Assessment

• • Quick fix with minimal changes, resolved CORS issue promptly.

Summary

Add scoped task page task/schedule switch

Allows users to toggle between tasks and scheduled tasks within agent or project pages, improving task organization and reducing clutter.

Health Assessment

• • Fast 2.8‑hour cycle with a single commit indicates a straightforward change with minimal review friction and low risk.

Summary

Sync release branch with latest canary

Aligns the weekly release branch with the latest canary, incorporating new UI test‑mock fixes and canary changes to improve build stability and feature parity.

Health Assessment

• • Fast merge with minimal review indicates a straightforward branch sync; the large line changes reflect many upstream commits but no new functionality was introduced.

Summary

Fix sub-agent bridge answer recovery

Ensures heterogeneous sub-agents correctly return their answers, preventing placeholder messages and improving reliability in queue mode.

Health Assessment

• • Quick fix with minimal rework, AI-assisted review accelerated the process.

Summary

Fix QStash deduplication ID encoding

Ensures QStash deduplication IDs are provider‑safe by encoding logical keys, preventing errors and preserving original identifiers.

Health Assessment

• • Quick fix with minimal changes, resolved provider error swiftly.

Summary

Upgrade pnpm setup action to v6

Updates GitHub Actions to use Node 24 runtime, eliminating deprecation warnings and aligning release workflows.

Health Assessment

• • Quick merge with minimal changes, indicating low complexity and high confidence.

Summary

Fix heterogeneous agent name scoping

Ensures consistent agent naming across personal, private, and shared workspaces, reducing confusion and improving user experience.

Health Assessment

• • Fast cycle time of 2 hours with only one review comment and two subsequent commits indicates a smooth, low‑friction fix.

Summary

Fix eval run topic title hints

This change removes unnecessary eval hints from run topic titles, improving clarity and consistency in evaluation logs.

Health Assessment

• • PR completed in 1.4 hours with minimal changes and no review, indicating a straightforward bug fix.

Summary

Fix connector ownership canonicalization

Ensures consistent connector ownership, preventing conflicts in discovery and authorization flows, improving reliability for users.

Health Assessment

• • Fast review and single rework cycle, but the change spans 20 files and 525 lines, indicating a substantial codebase impact.

Summary

Fix connect-agent device loading state flash

Eliminate UI flash by showing a skeleton while device data loads, keeping cached devices visible, and improving command-line hint formatting for a smoother user experience.

Health Assessment

• • Merged within minutes with a single commit, indicating a straightforward bug fix with minimal review friction.

Summary

Add workspace document comment support

Enables threaded comments with mentions, attachments, and rich‑text editing for workspace documents, improving collaboration and communication.

Health Assessment

• • Rapid 16‑hour cycle driven by automated Codex reviews, but extensive changes required multiple iterations and a large code footprint.

Summary

Stub base-ui ActionIcon for unit tests

Adds stubs for base‑ui components and MotionProvider to enable unit tests to run without app‑level dependencies, improving test reliability.

Health Assessment

• • Fast 1.5‑hour cycle with a single commit and no review friction; test stubs added to eliminate runtime failures in unit tests.

Summary

Add dedicated skeletons for lazy routes

Provides appropriate loading placeholders for each lazy‑loaded route, reducing layout shifts and improving user experience.

Health Assessment

• • Rapid AI‑assisted iteration produced a concise, well‑tested PR that added UI skeletons across many routes, improving UX with minimal review friction.

Summary

Center skeleton hero for consistent layout

Aligns the hero component in the share topic page to avoid layout shift, improving user experience.

Health Assessment

• • Quick UI fix with minimal changes, merged within 1.3 hours, no tests added.

Summary

Refactor: migrate lobe-cron to lobe-task

Migrates the internal cron job system to a new task scheduler, reducing technical debt and enabling easier scaling of background jobs.

Health Assessment

• • Fast review cycle with automated AI review, minimal code churn, and no new features introduced.

Summary

Add PR-time verify workflow for share micro app

Adds a CI workflow that validates the share micro app build, enforces size limits, and provides PR previews, reducing deployment risk and ensuring build stability.

Health Assessment

• • The PR was reviewed and merged within an hour, indicating a smooth process with minimal friction.

Summary

Update desktop CI to use variable overlay repo

Replaces hardcoded overlay repository references in desktop CI workflows with a variable, removing dependency on org‑transfer redirect and improving build reliability.

Health Assessment

• • Fast merge (0.1h cycle time) with minimal changes; no tests added; straightforward variable update.

Summary

Align toast radius and offsets with desktop layout container

Adjusts toast styling to match desktop layout container, improving visual consistency across desktop views.

Health Assessment

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

Summary

Health Assessment

Summary

Fix chat output cost estimate ratio

Improves cost estimation accuracy for chat outputs, ensuring billing aligns with actual usage and maintaining system caps.

Health Assessment

• • Quick fix with minimal changes, no review needed, merged within 1.1 hours.

Summary

Add workspace document comment schema

Adds persistent storage for workspace document comments, enabling future comment features and improving collaboration.

Health Assessment

• • Large migration added over 33k lines, reviewed within 11 hours, and required a few follow‑up commits to address automated review feedback, indicating moderate complexity but high impact on the database layer.

Summary

Add Grok provider bindings support

Enables Grok Build to use LobeHub provider bindings, expanding compatibility with OpenAI and Anthropic providers and improving flexibility and security.

Health Assessment

• • Quick review and merge within 3.3 hours, minimal rework, indicating a well-scoped feature addition.

Summary

Refactor UI components to new base-ui subpath

Migrates deprecated antd-based UI wrappers to @lobehub/ui/base-ui, improving component consistency and eliminating runtime warnings.

Health Assessment

• • Fast review cycle (15h) with AI-assisted review comments, minimal rework, and no critical bugs reported.

Summary

Align recommendation icon sizes on Home

Ensures consistent icon sizing across recommendation cards, improving UI consistency and user experience.

Health Assessment

• • Fast cycle time of 1.4 hours, minimal changes, no review needed, indicating a straightforward bug fix.

Summary

Add Kimi Code provider binding

Adds Kimi Code as a provider‑bound heterogeneous agent, enabling secure integration with OpenAI and Anthropic providers while keeping API keys isolated.

Health Assessment

• • The PR was reviewed and merged within 17 hours with minimal back‑and‑forth, indicating a smooth process despite the large code addition.

Summary

Enable disabled models from chat input

Adds an enable action for disabled chat models, enforces permission checks, and improves recovery feedback for users.

Health Assessment

• • Fast cycle time (13h) with minimal review comments, indicating straightforward bug fix and efficient automated review.

Summary

Remove repository setup scripts for Orb lifecycle

Centralizes Orb setup in Amp Project Settings, reducing repository clutter and simplifying maintenance.

Health Assessment

• • Fast merge with minimal changes; no tests required; low risk to production.

Summary

Align approval workflows with web

Enables consistent approval processes across web and agent interfaces, improving user experience and reducing errors.

Health Assessment

• • The PR received a quick automated review but required nine additional commits, indicating complex changes and extensive testing. The large code churn and multiple iterations suggest a high-risk, high-impact feature.

Summary

Add Pi provider bindings for desktop

Enables LobeHub to integrate Pi heterogeneous agent with OpenAI, Anthropic, and Google providers, improving flexibility and security by managing credentials internally.

Health Assessment

• • Merged within 16 hours with no reported review delays, indicating a smooth approval process and high confidence in the implementation.

Summary

Persist generic agent interventions in database

Adds robust persistence for agent interventions, enabling atomic approvals, retries, and marketplace ledger, improving reliability and consistency for the platform.

Health Assessment

• • Extensive database changes required multiple iterations; long cycle time indicates high complexity and potential integration risk.

Summary

Flatten quota calendar window list and cost

Improves UI clarity by consolidating window rows and displaying daily cost upfront, reducing visual clutter and enhancing user comprehension.

Health Assessment

• • Rapid review and merge with minimal iterations indicates a straightforward UI tweak with low risk.

Summary

Fix Kanban column header skeleton during grouping

Ensures consistent loading UI for Kanban columns, preventing stale labels during data fetch and improving user experience.

Health Assessment

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