The previous fix (513982b16) only added --external mupdf to the
package.json build:binary script, but the CI release job uses its
own inline build_binary function that was missing the flag. This
caused the release job to fail on tag push v13.17.3 with a
top-level await error from mupdf via markit-ai.
- Added support for writing to archive entries using `archive.ext:path/inside/archive` syntax for .tar, .tar.gz, .tgz, and .zip formats.
- Added ability to create new archives automatically when writing to non-existent archive subpaths with nested directory support.
- Implemented archive path resolution and validation with fallback to plain file writes for non-archive filenames.
- Added comprehensive test coverage for archive write operations including entry updates, archive creation, and fallback behavior.
- Removed 3 test cases from gh.test.ts that relied on implementation details and external state.
- Removed unused helper functions getCurrentHeadSha() and GhToolDetails import.
- Updated test assertions to use path.resolve() and corrected expected remote name from fork-contrib to forksrc.
- Reformatted else blocks in shell.rs for consistency.
- Replaced Python-based markitdown CLI with native markit-ai library for document and notebook conversion.
- Added support for converting Jupyter notebooks (.ipynb) to markdown via markit-ai integration.
- Created markit utility module with file/buffer conversion wrappers and standardized error handling.
- Removed markitdown from Python tools manager and eliminated external CLI dependency.
- Updated fetch, read, and scraper tools to use new markit conversion API across codebase.
- Added test coverage for ipynb file conversion through markit utility.
- Replaced blocking Tokio worker with non-blocking async I/O for pipe reading to prevent stalling the runtime when child processes keep pipes open.
- Refactored background job termination to use process group and process tree APIs instead of deprecated signal parsing, improving reliability across Unix and Windows.
- Added non-blocking pipe registration and fcntl-based flag management for Unix platforms to enable efficient async I/O without blocking workers.
- Added test case verifying read_output cancellation before pipe EOF to ensure proper cleanup behavior.
- Added archive reading support for .tar, .tar.gz, .tgz, and .zip files with virtual subpath syntax (archive.ext:path/inside/archive).
- Added archive listing and subdirectory navigation capabilities to explore archive contents without extraction.
- Added /tools slash command to display visible agent tools and their capabilities.
- Implemented ArchiveReader module with openArchive(), parseArchivePathCandidates(), and archive content extraction methods.
- Added fflate 0.8.2 dependency for archive decompression support.
- Added comprehensive test coverage for TAR, TAR.GZ, TGZ, and ZIP archive reading with fixture generation utilities.
- Added `/tools` slash command to display agent-visible tools in interactive sessions.
- Added `handleToolsCommand()` method to InteractiveMode to delegate tools display to command controller.
- Added `buildToolsMarkdown()` utility function to generate formatted markdown table of available tools.
- Refactored `showMarkdownPanel()` helper to reduce duplication in markdown display logic across commands.
- abortableSleep used Bun.sleep which vitest fake timers do not intercept;
switch to setTimeout via Promise.withResolvers so vi.useFakeTimers() +
vi.advanceTimersByTime() work correctly in the polling loop tests
- getCurrentHeadSha() and GhRunWatchTool session cwd used a hardcoded
/work/pi path; replace with import.meta.dir so the tests resolve the
real repo HEAD on any machine
- add timeout-minutes: 30 to the CI test job as a safety net against
future hangs blocking release runs
- Fixed artifact storage for non-persistent sessions to use in-memory fallback instead of returning undefined, enabling proper spill truncation for all session types.
- Added in-memory artifact map and counter to SessionManager for fallback storage when file-based ArtifactManager is unavailable.
- Updated saveArtifact() JSDoc to reflect new behavior of returning artifact IDs for all sessions.
- Enhanced `gh_run_watch` tool call rendering to display animated spinner status and target description (run ID, branch, or current HEAD) with improved visual hierarchy.
- Refactored renderer to use `formatStatusIcon` helper for consistent status display across spinner frames.
- Added gh-renderer.ts module with GitHub Actions workflow visualization for terminal output.
- Enhanced gh_pr_view tool to fetch and display inline review comments alongside pull request reviews.
- Improved gh_run_watch tool output with structured job state tracking and failed log formatting.
- Fixed gh_run_watch to resolve explicit branch watches against GitHub API branch head instead of local HEAD.
- Fixed gh_* tool outputs to spill full large results to artifacts instead of pre-truncating.
- Standardized trailing newlines across 100+ prompt template files for POSIX compliance.
- Added gh_pr_checkout tool to check out GitHub pull requests into git worktrees with branch metadata.
- Added gh_pr_push tool to push checked-out pull request branches to source fork with force-with-lease support.
- Implemented 15+ git helper functions for worktree, branch, and remote management operations.
- Extended GhPrViewData interface with headRefOid, headRepository, isCrossRepository, and maintainerCanModify fields.
- Added comprehensive integration tests for gh_pr_checkout and gh_pr_push tools with fixture setup and cleanup.
sanitizeRehydratedOpenAIResponsesAssistantMessage now removes the
assistant providerPayload for Responses-family messages, not just the
stale thinkingSignature. After rehydration the native replay snapshot
belongs to a previous live provider connection and replaying it on a
warmed session causes 401 rejections from GitHub Copilot.
User/developer providerPayload is preserved by the caller since it
carries durable compaction and preserved history needed for valid cold
request reconstruction.
Fixes#592
* fix(ai): update spoofed Gemini CLI User-Agent to v0.35.3 format
Gemini CLI v0.35+ changed its User-Agent format:
- Dropped 'google-api-nodejs-client/10.5.0' suffix
- Added surface field (third param in parens)
- New format: GeminiCLI/VERSION/MODEL (PLATFORM; ARCH; SURFACE)
The stale v0.34.0 UA was contributing to google-gemini-cli/gemini-3.1-pro-preview
being rejected, alongside Google's new abuse detection for third-party CCA usage
(google-gemini/gemini-cli#22970).
Verified: model responds successfully with updated UA.
* feat(scripts): add spoofed version drift checker
Fetches latest stable release from google-gemini/gemini-cli GitHub
releases and compares against the hardcoded version in the provider.
bun check-spoofed-versions # report drift, exit 1 on mismatch
bun check-spoofed-versions --update # apply version bump in-place
Extensible for additional spoofed tools (Antigravity, Copilot headers).
- Replaced truncation with line wrapping for diff output in the edit tool renderer.
- Added `wrapEditRendererLine` to preserve gutter markers and line numbers on continuation lines.
Fixes#578
- Applied `disabledProviders` setting to `getAvailable()` and `getDiscoverableProviders()` so `--list-models` and `/model` hide models from disabled providers.
- Initialized settings before model refresh in the `--list-models` code path.
- Added tests verifying disabled providers are excluded from both methods.
Fixes#588
- Added `thinkingSignature` field to thinking content blocks to preserve the original reasoning field name (e.g., `reasoning_text`, `reasoning_content`) for accurate follow-up requests.
- Enhanced thinking block handling to track and distinguish between different reasoning field types, enabling proper field name preservation across multiple conversation turns.
- Updated `appendThinking` function to accept optional signature parameter and conditionally create new thinking blocks when signature changes.
- Added test case validating that streamed reasoning field names are preserved for follow-up requests.
- Added first-event timeout detection for streaming responses to prevent stuck requests and enable faster recovery from stream stalls.
- Added PI_STREAM_FIRST_EVENT_TIMEOUT_MS environment variable to configure first-event timeout behavior independently from idle timeouts.
- Fixed Anthropic stream timeout errors to be properly retried with first-event recognition, distinguishing first-event from idle timeouts.
- Enhanced idle-iterator utility with FirstEventWatchdog mechanism and markFirstStreamEvent wrapper for unified timeout handling across OpenAI, Azure OpenAI, and Anthropic providers.
- Removed code_search tool and all related Exa API integration from web search module.
- Deleted code-search.ts implementation including CodeSearchResponse types and executeCodeSearch() function.
- Updated getSearchTools() to return only webSearchCustomTool, removing dual-tool support.
- Removed code_search tool documentation from README and prompt files.
can1357