The official component-detection-manifest schema uses camelCase keys.
This file used PascalCase (e.g. "Registrations", "Component"), which
caused the OSS license tool to crash silently.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Update rate limit message
* Improve rate limit messages for better user clarity
* removing 'ed' in case not fixed on server
* nvm, it was fixed
---------
Co-authored-by: SteVen Batten <sbatten@microsoft.com>
Cache breakpoints in the summarization prompt cause the prompt-tsx
budget trimmer to throw BudgetExceededError ('No lowest priority node
found') when all remaining children contain cache points. The trimmer
resets its eviction candidate each time it encounters a cache point,
and when every child has one, it never finds a node to evict.
These breakpoints were already stripped from the output messages by
stripCacheBreakpoints() before sending to the model, so they served
no purpose in the summarization render.
Fix: pass enableCacheBreakpoints: false when rendering
ConversationHistorySummarizationPrompt.
The reasoningEffort telemetry property on response.success and
response.error events was only reading from requestBody.reasoning?.effort
(Responses API / OpenAI), missing Claude models which set effort via
requestBody.output_config?.effort (Messages API). This adds a fallback
to output_config.effort so both events capture the reasoning effort
level regardless of which API path is used.
Co-authored-by: Kevin Kent <kevinkent@microsoft.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* NES: enable trigger on active editor change by default
- Set triggerOnEditorChangeAfterSeconds default to 10 (was undefined)
- Set triggerOnEditorChangeStrategy default to AfterAcceptance (was Always)
- Replace unconditional same-line cooldown bypass with smart reset:
cooldown is enforced within a file session but clears when switching
away and returning, so NES re-triggers on the same line after a
round-trip to another file
- Update tests to match new defaults and behavior
* fix indentation
xtab: replace getFetchFailure polling with stream-level error propagation (#308745)
Add FetchStreamError class and reject() method to FetchStreamSource so fetch
failures propagate through the async iterable stream. Consumers now
catch errors naturally via for-await instead of polling a mutable
closure at ~7 call sites.
- Added FetchStreamError (carries NoNextEditReason) in xtab/common,
keeping platform/chat free of inlineEdits dependencies
- Added FetchStreamSource.reject() delegating to AsyncIterableSource
- Updated _performFetch to call reject()/resolve() instead of setting
chatResponseFailure
- Removed getFetchFailure parameter from all response format handlers
- Added FetchStreamError catch in _streamEditsImpl and
xtabCustomDiffPatchResponseHandler
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* Bump version to 1.117.0
* npm i
* wait to do engine version bump
* Revert "wait to do engine version bump"
This reverts commit 9db1c0feb6.
* Add Copilot extension tests to Linux/Windows Electron integration test runs
* Remove failing step that we moved to the main build
---------
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Previously rejected NES entries now show a 👎 icon instead of no
specific icon. Uses direct outcome assignment to avoid false warnings
on the succeeded→previouslyRejected transition. Also registers
the rebasedEdit (when present) with the rejection collector.
* Agents - switch back to using repository state in the session tracker to avoid a race condition
* Pull request feedback
* Add initial update of the state when a session is opened
Hooks can be disabled... we don't want the user to disable hooks and then key functionality goes away.
* Edit tracking
* Enter/Exit plan mode
* tool logging
all now go through the message dispatching instead of via hooks.
Additionally, there was duplicate logic in the hooks world for logging so that was all removed since we already log hooks in the dispatcher.
this allowed us to completely delete the hook registery code.
* Cleanup code related to the repository state
* Remove code specific to the Agents app
* HasGitRemote should be computed at the end of each turn to account for sessions that were started before the metadata was added
* nes: implement cursor line distance check for serving from cache
* address review: use rebasedEdit for rejection, skip cross-file cursorOffset, fix test types
nes: fix: address PR review for _performFetch extraction
- Rename FetchResult.Error to FetchResult.FetchFailure to avoid
confusion with built-in Error class
- Move fetchResultPromise .then/.catch/.finally handlers before
await Promise.race so early-return paths get setFetchEndTime()
telemetry
