Pull Request Explorer

Summary

Fix UI flip‑flop on rapid automation toggles

This patch stabilizes the task automation UI, preventing visual jitter and data race issues when users rapidly switch between scheduled and heartbeat modes, improving reliability and user experience.

Health Assessment

• • Fast review and single commit indicate a straightforward bug fix with minimal risk.

Summary

Fix bot channel functionality

Corrects bot channel behavior and adds documentation, improving user experience and reliability.

Health Assessment

• • Fast cycle time and minimal review comments indicate smooth integration, but the large code change suggests careful testing and monitoring.

Summary

Health Assessment

Summary

Fix tool forwarding into compression budget

Ensures large tool manifests are counted during LLM calls, preventing context window overflows and improving reliability.

Health Assessment

• • Quick, single‑commit fix with automated review bots; minimal rework and no major blockers.

Summary

Show cloud credentials alert and disable input

Adds a warning banner and disables chat input when cloud credentials are missing, improving user experience and preventing errors.

Health Assessment

• • Fast cycle time and minimal review comments indicate a smooth, low-risk change.

Summary

Health Assessment

Summary

Fix pre‑flight context check for OpenAI‑compatible providers

Prevents doomed requests to upstream by validating context size before sending, improving reliability for NVIDIA and DeepSeek providers.

Health Assessment

• • Fast cycle time and single commit indicate a straightforward, low‑risk bug fix with minimal review friction.

Summary

Fix subagent event leakage in gateway handler

Prevents subagent tool events from leaking into the main gateway, eliminating UI inconsistencies and unnecessary processing.

Health Assessment

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

Summary

Rename chat mode label to '智能' in zh-CN

Updates Chinese locale to rename the chat mode label from '智能体' to '智能', improving clarity and consistency for Chinese users.

Health Assessment

• • Quick, single-commit PR with no review friction, indicating a straightforward i18n update.

Summary

Fix file storage upload serialization

Ensures file storage quota checks and file creation occur within a single database transaction, preventing race conditions during concurrent uploads.

Health Assessment

• • Quick review and merge with minimal changes, indicating a straightforward bug fix.

Summary

Fix Windows CLI shim resolution for agents

Ensures stable agent startup on Windows by resolving npm‑generated CLI shims, improving reliability for desktop users.

Health Assessment

• • Fast review with minimal back‑and‑forth, but the change is large, affecting over 500 lines of TypeScript code to resolve Windows CLI spawning issues.

Summary

Fix agent profile mention list to include hidden builtin tools

Ensures hidden but discoverable built‑in tools appear in the system prompt editor’s @‑mention dropdown, improving agent configuration accuracy.

Health Assessment

• • Quick fix with minimal changes, AI‑assisted code generation, no blockers.

Summary

Restore legacy .json fallback for agent tracing

Ensures older trace snapshots are accessible after compression rollout, preventing 404 errors and maintaining trace visibility for pre‑rollout operations.

Health Assessment

• • Quick AI‑generated fix with minimal changes, fast review and merge, low risk.

Summary

Update subscription credit top-up copy

Adds updated user-facing copy for subscription credit top-up and auto top-up guidance, improving clarity for Free plan users.

Health Assessment

• • Rapid review and merge with minimal changes; no tests required, indicating low complexity and risk.

Summary

Fix MiniMax max_tokens overflow

Prevents context window errors by capping max_tokens based on model context, improving reliability for large tool definitions.

Health Assessment

• • Quick 2‑hour cycle with no review rounds indicates a straightforward, low‑risk bug fix; the large code addition is well‑contained within the provider logic.

Summary

Fix submenu close behavior in ActionDropdown

Fixes submenu persistence issue in chat input dropdown, improving user experience by ensuring submenus close when another opens or focus leaves.

Health Assessment

• • Rapid resolution with single commit and immediate review, indicating low complexity and minimal risk.

Summary

Fix context engine token accounting for tool calls and reasoning

Improves token counting to correctly trigger compression, preventing context window overflows and ensuring reliable agent conversations.

Health Assessment

• • PR resolved a critical token accounting bug in under an hour with minimal review friction, leveraging AI-generated code. The large code change (over 650 lines) was merged quickly, indicating confidence in the fix but also a moderate risk due to the scope of the update.

Summary

Add onBeforeCall and onAfterCall lifecycle hooks

Enables server-side tools to invalidate UI caches after mutations, improving user experience.

Health Assessment

• • The PR had a large scope with 9 files changed, but the review process was relatively smooth with a short time to first review.

Summary

Restore HTML preview for AssistantGroup messages

Fixes rendering of full HTML documents in grouped assistant messages, enabling iframe preview and expand drawer functionality.

Health Assessment

• • Merged within 30 minutes with a single commit, indicating a straightforward bug fix with minimal review friction.

Summary

Compress agent snapshots with zstd for S3

Compresses agent operation snapshots using zstd before uploading to S3, cutting storage costs by ~88% and improving upload performance.

Health Assessment

• • Fast cycle time and minimal review friction indicate a straightforward performance improvement with low risk.

Summary

Fix diagnostic context for PlaceholderVariablesProcessor errors

Improves error diagnostics in the context engine, enabling clearer debugging and better user support.

Health Assessment

• • Rapid review and multiple commits indicate active rework but minimal friction; large code changes suggest high impact on the core engine.

Summary

Fix parallel tool count rollback in heterogeneous agent

Ensures consistent tool count during parallel tool execution, improving reliability of agent responses.

Health Assessment

• • Fast review and single commit indicate minimal friction; the change is small to medium in scope and does not affect public APIs, keeping risk low.

Summary

Update i18n translations

Adds updated translations for multiple locales, ensuring consistent user experience across languages.

Health Assessment

• • Automated i18n update with minimal review, merged after bot comment; long cycle time due to scheduling but low risk.

Summary

Pin Vite to 8.0.12 to avoid regression

This PR pins Vite to version 8.0.12 in both package.json files to prevent a production‑breaking regression caused by Vite 8.0.13’s updated rolldown dependency. This ensures stable tRPC calls and application startup.

Health Assessment

• • Fast merge with minimal changes, no review friction, low risk.

Summary

Add bot watch support

Enables bots to watch keyword instructions, expanding messenger integration for users.

Health Assessment

• • Fast cycle time and immediate review indicate high team efficiency. Large scope but minimal review friction suggests confidence in code quality.

Summary

Fix onboarding agent route stall and redirect

Improves onboarding flow reliability by preventing stale redirects and eliminating loading stalls, ensuring a smoother user experience.

Health Assessment

• • Quick resolution with minimal iterations; AI-assisted review expedited the process.

Summary

Add chat mode support and redesign input bar

Enhances user experience by introducing chat mode and a streamlined input interface, reducing token usage and simplifying interactions.

Health Assessment

• • The PR was merged within 24 hours, indicating a fast review cycle, but the large number of commits after the initial review and the extensive code changes suggest moderate friction and a high risk of integration issues.

Summary

Fix resume request trigger preservation

Ensures tool interaction resumes maintain request context, improving continuity and user experience.

Health Assessment

• • Fast turnaround with minimal review, indicating a straightforward bug fix.

Summary

Fix task template skeleton layout shift

Reduces layout shift when users refresh task template recommendations by matching the loading placeholder to the final card layout.

Health Assessment

• • The PR was reviewed quickly with a cycle time of 7.2 hours and had multiple commits and reviews, indicating active collaboration and testing.

Summary

Fix desktop local file preview restrictions

Prevents unauthorized local file access and improves security for desktop previews.

Health Assessment

• • Rapid review and merge within 5 hours indicates efficient process; large code changes but minimal review rounds suggest confidence in the changes.

Summary

Simplify and refresh skill docs

Clean up skill documentation by removing redundant auto‑trigger text and updating directory listings, improving clarity for users.

Health Assessment

• • Merged within 0.9 hours after a single review comment, indicating a straightforward, low‑risk documentation update.

Summary

Add Visual Media Inspector and HTML Preview Refactor

Adds a visual media inspector, streamlines portal HTML rendering, and deduplicates context engine traces, improving user experience and system performance.

Health Assessment

• • Fast turnaround with minimal review iterations; AI assistance accelerated development.

Summary

Add benchmark agent config parser

Adds a parser for benchmark agent memory extraction configuration, enabling more flexible memory handling.

Health Assessment

• • Quick review and merge within 12 hours, single commit, minimal scope, low risk.

Summary

Add source ids to extraction schemas

Support for source ids in extraction schemas to improve data handling.

Health Assessment

• • Fast review and merge, low risk changes

Summary

Fix system bot error

Corrects a system bot error that caused failures in messaging, improving user experience and reducing support tickets.

Health Assessment

• • Rapid turnaround with minimal review, indicating a straightforward bug fix; large code churn reflects extensive i18n updates across multiple locales.

Summary

Fix indent for heterogeneous agent topic rows

Restores consistent indentation for agent topic rows, ensuring visual alignment across agents.

Health Assessment

• • Quick fix with minimal changes, fast review, no rework.

Summary

Fix desktop runtime externals split

Improves Electron packaging by separating runtime externals from native dependencies, reducing duplicate handlers and ensuring proper runtime resolution.

Health Assessment

• • Quick review and merge within 45 minutes, indicating a straightforward bug fix with minimal rework.

Summary

Fix agent signal blocking during execution

Prevents agent execution from being blocked by signal enqueue failures, improving reliability.

Health Assessment

• • Quick fix with minimal changes, no rework, fast review.

Summary

Health Assessment

Summary

Add onboarding request trigger via metadata

Enables distinguishing onboarding chat traffic for routing, billing, and logs, improving billing accuracy and onboarding flow.

Health Assessment

• • Rapid review and merge with minimal iterations, indicating clear scope and high confidence.

Summary

Refactor agent signal and self‑iteration logic

Improves stability and clarity of agent signal handling, reducing errors in self‑iteration and lab features.

Health Assessment

• • Fast cycle time and minimal review iterations indicate an efficient refactor with clear scope.

Summary

Fix memory embedding input limit

Limits memory embedding input to prevent token overrun, ensuring reliable memory queries and reducing errors.

Health Assessment

• • Quick turnaround with minimal review, indicating a straightforward bug fix.

Summary

Fix file preservation on NoSuchKey errors

Prevents data loss by ensuring files are not deleted when storage errors occur, and cleans orphan database records, improving reliability for users.

Health Assessment

• • The PR resolved a critical data‑loss risk but had a long cycle time, suggesting extended testing or scheduling delays.

Summary

Polish SideBarDrawer and header layout

Improves UI consistency and spacing in the navigation panel, enhancing user experience.

Health Assessment

• • The PR was reviewed and merged within 2.7 hours with minimal changes, indicating a smooth, low-risk process.

Summary

Add Agent Signal marker to receipt descriptions

Enhances receipt UI by embedding the Agent Signal indicator within the description line, preserving primary icons and adding a shared label for clearer user experience.

Health Assessment

• • Fast merge with minimal review, indicating a straightforward UI improvement with low risk.

Summary

Hide ModeSwitch in production environment

Fixes the ModeSwitch component rendering in production by wrapping it with an isDev guard.

Health Assessment

• • Quick review and merge with no significant rework or discussion.

Summary

Clarify sandbox TTL and add public-repo fork push guide

Improves cloud context for sandbox runs by clarifying the TTL and providing a guide for pushing to public repositories, reducing failures when the system lacks write access.

Health Assessment

• • Fast cycle time (0.1h) with no rework and minimal comments indicates a straightforward maintenance change with low risk.

Summary

chore: translate non-English comments to English

Translates comments in file-loader tests to English, improving code readability and maintainability.

Health Assessment

• • Fast turnaround with minimal review, indicating straightforward change.

Summary

Fix memory retrieval sufficiency rules

Clarify when retrieved memories are sufficient for answering or taking memory actions, improving the memory tool's functionality.

Health Assessment

• • The PR was merged quickly, indicating a straightforward fix.

Summary

Online Messager

Adds internationalization support for the online messenger feature, updating locale files to improve global user experience.

Health Assessment

• • Fast cycle time and minimal review iterations indicate a straightforward i18n update with low complexity.