Update @github/copilot to 1.0.56#1504
Merged
Merged
Conversation
- Updated nodejs and test harness dependencies - Re-ran code generators - Formatted generated code
Contributor
There was a problem hiding this comment.
Pull request overview
This PR updates the vendored @github/copilot dependency from the prerelease 1.0.56-2 to 1.0.56 and refreshes generated SDK documentation derived from the updated schema.
Changes:
- Updated Node.js and test harness package manifests/lockfiles to
@github/copilot^1.0.56. - Refreshed generated RPC documentation for context limits and billing token budget fields across several SDKs.
- Updated sample lockfile metadata for the local Node.js SDK dependency.
Show a summary per file
| File | Description |
|---|---|
test/harness/package.json |
Updates harness @github/copilot dev dependency. |
test/harness/package-lock.json |
Locks harness dependency and platform packages to 1.0.56. |
rust/src/generated/api_types.rs |
Refreshes generated Rust RPC field documentation. |
python/copilot/generated/rpc.py |
Refreshes generated Python RPC field documentation. |
nodejs/src/generated/rpc.ts |
Refreshes generated TypeScript RPC field documentation. |
nodejs/samples/package-lock.json |
Updates sample lock metadata for the local SDK package dependency. |
nodejs/package.json |
Updates Node.js SDK @github/copilot dependency. |
nodejs/package-lock.json |
Locks Node.js SDK dependency and platform packages to 1.0.56. |
go/rpc/zrpc.go |
Refreshes generated Go RPC field documentation. |
dotnet/src/Generated/Rpc.cs |
Refreshes generated .NET RPC field documentation. |
Copilot's findings
Files not reviewed (4)
- go/rpc/zrpc.go: Language not supported
- nodejs/package-lock.json: Language not supported
- nodejs/samples/package-lock.json: Language not supported
- test/harness/package-lock.json: Language not supported
- Files reviewed: 2/10 changed files
- Comments generated: 0
Runtime 1.0.56 changed disconnect semantics so that ForceStop of the last RPC owner now triggers session cleanup. The previous warm-only test for `Should_Keep_Pending_External_Tool_Handleable_On_*_Resume_When_ContinuePendingWork_Is_False` asserted SessionWasActive == true and that `handlePendingToolCall` fed a result into the assistant reply -- assumptions that only hold for warm resume. On cold resume the runtime intentionally auto-completes orphan tool calls with a synthetic interrupt result, so the SDK call correctly returns success=false. Split that test into two scenarios across Node, Go, .NET, and Python: - Warm (original client stays connected): SessionWasActive=true, handlePendingToolCall returns success=true, and the assistant echoes the supplied result. - Cold (original client ForceStopped before resume): SessionWasActive=false, handlePendingToolCall returns success=false, and a follow-up turn confirms the resumed session is still healthy. In warm mode the resumed client must not re-register the external tool (it would clash with the original owner); in cold mode it re-registers with a throwing handler to assert the runtime does not re-invoke the handler on resume. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This comment has been minimized.
This comment has been minimized.
Runtime PR #9040 (commit b8e1220b45) made SDKServer.handleConnectionClosed
clean up session state when the last RPC owner disconnects. As a result,
the existing Should_Continue_Pending_{Permission,External_Tool}_Request_
After_Resume tests, which ForceStop the suspended client and then expect
the resumed client to successfully complete the pending request, no longer
pass: HandlePendingPermissionRequest/HandlePendingToolCall return
success=false after the cold cleanup runs.
A runtime-side fix is being tracked separately. Skip these two tests in
all four SDKs (.NET, Go, Node, Python) with a TODO referencing the
runtime contract change so they can be re-enabled once that work lands.
Also drop the GetFinalAssistantMessageAsync continuation assertions from
the .NET and Go warm theory and from the older OLD tests, matching the
Node and Python tests which never made that assertion. The
ContinuePendingWork=false warm path provides no guarantee that the
suspended client's agentic loop will propagate a final assistant.message
to the resumed client, and waiting for one was flaking on Linux .NET CI.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This comment has been minimized.
This comment has been minimized.
The previous wording ('Pending runtime fix') implied a runtime change was being tracked separately. That isn't accurate: runtime 1.0.56 (copilot-agent-runtime PR #9040) intentionally cleans up the session when the last RPC client disconnects, and there is no in-flight fix planned.
The OLD tests model same-process ForceStop+resume and rely on the old behavior where the session stayed alive in the runtime process. They need to be redesigned to either keep an owner connected (warm resume) or to model true process-restart resume from persisted session state.
No code change beyond the skip-reason text; tests remain skipped while the redesign is figured out.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Contributor
Author
Cross-SDK Consistency ReviewThis PR updates the
|
| File | Stale comment |
|---|---|
ModelBillingTokenPrices.java |
/** Maximum context window tokens for the default tier */ |
ModelBillingTokenPricesLongContext.java |
/** Maximum context window tokens for the long context tier */ |
SessionMetadataContextInfoResult.java |
/** Total context limit for /context display. promptTokenLimit + min(32k or 64k, outputTokenLimit) depending on model. */ |
Other SDKs now read:
- "Prompt token budget (max_prompt_tokens) for the default/long context tier. The total context window is this value plus the model's max_output_tokens."
- "Total context limit for /context display: promptTokenLimit + outputTokenLimit (the model's full max_output_tokens reserved on top of the prompt budget)."
To regenerate Java, run: cd java && mvn generate-sources -Pcodegen
i️ Java pending_work_resume E2E tests
The PR updates pending_work_resume E2E tests in go, dotnet, nodejs, and python. Java has no equivalent PendingWorkResume E2E tests yet — this is a pre-existing gap (not introduced by this PR), but worth noting for future parity work.
Generated by SDK Consistency Review Agent for issue #1504 · ● 7.2M · ◷
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Automated update of
@github/copilotto version1.0.56.Changes
@github/copilotinnodejs/package.jsonandtest/harness/package.jsonscripts/codegen)Next steps
When ready, click Ready for review to trigger CI checks.