Pull Request Explorer

Summary

Fix custom cache handlers missing on initial requests

Ensures that custom cache handlers are correctly applied on the first request, preventing fallback to built‑in handlers and improving cache consistency and performance.

Health Assessment

• • Fast turnaround with a single commit and no major rework; the change is small and focused on a specific runtime bug.

Summary

Enable Draft Mode for Unlisted Routes on Adapter Deployments

This change ensures preview mode works correctly for unpublished URLs on adapter deployments, improving preview reliability for developers and content editors.

Health Assessment

• • Fast cycle time and minimal rework indicate a straightforward bug fix with clear test coverage.

Summary

Prevent chat resumption after terminal text

Fixes UI bug where completed tool calls followed by terminal error text caused automatic resumption, improving user experience and preventing unintended requests. Ensures chat helper correctly distinguishes completed tool output from subsequent error messages.

Health Assessment

• • Quick resolution with minimal rework; PR merged within 2.4 hours, indicating efficient review and testing.

Summary

Fix agent UI stream persistence with originalMessages

Allows agent UI streams to run persistence flows when only originalMessages are provided, improving reliability and preventing execution failures.

Health Assessment

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

Summary

Add optional question to gateway fallback conditions

Allows gateway fallback conditions to be defined without a question, enabling dynamic policy evaluation. This reduces configuration complexity and supports use cases where questions are generated at runtime.

Health Assessment

• • Fast review and merge (1.6h cycle time) indicate low complexity and high confidence in the change.

Summary

Fix bundle analyzer source table delta sorting

Improves sorting accuracy and user experience in the bundle analyzer by correctly ordering delta changes and applying locale-aware tie-breaking, ensuring developers see accurate change metrics.

Health Assessment

• • Resolved quickly with minimal changes and fast review, indicating low complexity and risk.

Summary

Version Packages

Automates publishing of updated package versions to npm, updating changelogs and package metadata.

Health Assessment

• • Fast merge with minimal review, typical release automation.

Summary

Automated package version release

Publishes updated package versions to npm automatically via Changesets, streamlining release workflow.

Health Assessment

• • Fast merge with minimal review, indicating a straightforward release process.

Summary

Version Packages

Automated release of multiple SDK packages, updating changelogs and package versions.

Health Assessment

• • Merged within 1 hour with no review comments, indicating a straightforward automated release.

Summary

Fix resumeStream UI message stream after disconnect

Ensures UI message streams resume correctly after network disconnects, improving user experience and reliability.

Health Assessment

• • Quick fix with minimal changes, no blockers, and fast review cycle.

Summary

Add Claude Sonnet 5.5 support to SDK

Adds support for Claude Sonnet 5.5 model, updating capabilities and examples, enabling new thinking settings for developers.

Health Assessment

• • Rapid review and merge (0.3h) indicates high confidence and minimal friction; the large code change was completed quickly with no blockers.

Summary

Add Claude Sonnet 5.5 support to SDK

Adds support for Claude Sonnet 5.5 model, enabling new thinking modes and updated options across Anthropic, Amazon Bedrock, Google Vertex, and Gateway providers.

Health Assessment

• • Rapid review and merge within 18 minutes indicates high confidence and minimal friction.

Summary

Add Claude Sonnet 5.5 model support

Adds support for Claude Sonnet 5.5, enabling new thinking modes and improved tool handling for the Anthropic provider, expanding the SDK’s capabilities for developers.

Health Assessment

• • Fast review and merge with minimal rework, indicating a straightforward implementation of the new model support.

Summary

Fix resumeStream UI message resumption after disconnect

The fix restores chat message continuity after network interruptions, ensuring users see complete messages. This improves reliability and user experience in AI-powered chat interfaces.

Health Assessment

• • Quick resolution with minimal code changes and no review back‑and‑forth, indicating a low‑risk, low‑friction fix.

Summary

Fix stale snapshots after rename

Updated e2e snapshots to reflect renamed unstable features, restoring test stability.

Health Assessment

• • Quick fix with immediate merge, minimal review and rework.

Summary

Normalize orphaned callers in pruned history

Fixes a bug that caused errors when pruning conversation history, ensuring reliable tool calls and preventing API failures for developers using the Anthropic SDK.

Health Assessment

• • Merged within 40 minutes with minimal review, indicating a straightforward bug fix with low risk to the codebase.

Summary

Isolate database setup to avoid test timeouts

Ensures database tests run reliably on CI, preventing timeouts that could delay deployments.

Health Assessment

• • Quick fix with minimal changes, resolved CI timeout issue.

Summary

Normalize orphaned callers in Anthropic history

Fixes tool call serialization errors in the Anthropic provider, ensuring correct pruning and warning emission for v6 SDK, improving reliability for developers.

Health Assessment

• • Merged within 12 minutes with no review comments, indicating a straightforward bug fix backed by extensive tests and documentation updates.

Summary

Fix tool metadata preservation in UI streams

Ensures tool execution metadata (status, cache, timing, download info) is correctly displayed in UI, improving transparency and debugging.

Health Assessment

• • Quick turnaround with minimal rework; PR resolved within 48 minutes and merged without major issues.

Summary

Escape untrusted terminal control characters

Prevents terminal injection attacks by sanitizing control characters in TUI output, ensuring safe display of agent messages and tool content. Maintains original content integrity while protecting users from malicious terminal sequences.

Health Assessment

• • The PR was merged within 1.1 hours, indicating efficient review and minimal friction. Comprehensive test coverage and regression checks ensured confidence in the fix.

Summary

Health Assessment

Summary

Fix response stream pipe handling on client disconnect

Ensures server streams are properly cancelled and resources freed when a client disconnects, preventing leaks and improving reliability.

Health Assessment

• • Quick backport fix with minimal changes, no blockers, fast review cycle.

Summary

Fix response stream cleanup on client disconnect

Prevents resource leaks and improves reliability when clients disconnect during streaming responses.

Health Assessment

• • Quick fix with minimal changes, fast review and merge, low risk

Summary

Add shared and separate harness sandbox examples

Provides clear examples of shared and separate harness sandboxes, enabling developers to better understand and implement sandbox integration for AI functions. This improves developer onboarding and reduces integration errors.

Health Assessment

• • F
• • a
• • s
• • t
• •
• • r
• • e
• • v
• • i
• • e
• • w
• •
• • a
• • n
• • d
• •
• • m
• • e
• • r
• • g
• • e
• •
• • w
• • i
• • t
• • h
• •
• • m
• • i
• • n
• • i
• • m
• • a
• • l
• •
• • c
• • h
• • a
• • n
• • g
• • e
• • s
• •
• • i
• • n
• • d
• • i
• • c
• • a
• • t
• • e
• • s
• •
• • l
• • o
• • w
• •
• • c
• • o
• • m
• • p
• • l
• • e
• • x
• • i
• • t
• • y
• •
• • a
• • n
• • d
• •
• • h
• • i
• • g
• • h
• •
• • c
• • o
• • n
• • f
• • i
• • d
• • e
• • n
• • c
• • e
• •
• • i
• • n
• •
• • t
• • h
• • e
• •
• • e
• • x
• • a
• • m
• • p
• • l
• • e
• • s
• • .