This repository was archived by the owner on Jun 28, 2026. It is now read-only.
chore: wire Windows GitHub stack test branch - #3
Closed
jamiechicago312 wants to merge 4 commits into
Closed
Conversation
Co-authored-by: openhands <openhands@all-hands.dev>
Co-authored-by: openhands <openhands@all-hands.dev>
Co-authored-by: openhands <openhands@all-hands.dev>
Co-authored-by: openhands <openhands@all-hands.dev>
jamiechicago312
pushed a commit
that referenced
this pull request
Jun 18, 2026
… (#1390) (OpenHands#1391) * fix: ACP model picker/chip gaps vs agent-server 1.28.1 Addresses four issues identified in #1390: 1. **Provider registry drift**: Add check-acp-providers-sync.mjs script to detect when the bundled @openhands/typescript-client ACP registry diverges from the targeted agent-server version (1.28.1 is missing 18 models, offering 2 not in registry). New acp-providers-sync.yml workflow runs on package.json changes and schedules weekly to catch upstream divergence. 2. **In-conversation model switch before first message**: Fix 409 handling in useSwitchAcpModel. Before the first message there is no ACP session yet, so the agent-server returns 409. Now gracefully falls back to persisting the model in agent settings (same as home-page path) instead of erroring. 3. **Native model string formatting**: Add formatNativeModelString() to strip routing prefixes (e.g. "anthropic/claude-sonnet-4-5" → "claude-sonnet-4-5") so OpenHands conversation chips show meaningful model names, consistent with ACP chip behavior. 4. **Stale comment**: Update comment in useChatInputModelState clarifying that cloud ACP conversations now persist llm_model as acp_model (per OpenHands #14510), not null. The fallback still works and the new behavior is correct. Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com> * fix: surface an honest error for pre-session ACP model switch The in-conversation 409 path previously fell back to persisting the model as the agent-settings default. For an already-created conversation that's a silent no-op: the agent's acp_model is baked at creation and session/set_model needs a live session, so the settings write only changes the *next* conversation's default, not this one — while the UI reports success. Surface a clear error instead (CHAT_INTERFACE$ACP_MODEL_SWITCH_REQUIRES_SESSION) telling the user to send a message first, then switch. The home-page path (conversationId === null) is unchanged — that legitimately sets the next conversation's default. Temporary until the SDK persists a pre-session switch (OpenHands/software-agent-sdk#3763), after which the 409 disappears. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * refactor: harden ACP 409 detection + relocate native-model formatter Follow-up cleanup on the ACP model-chip/switch fixes: - use-switch-acp-model: detect the pre-session 409 via the shared getErrorStatus() instead of a hand-rolled `error.status` check. The cloud backend routes through axios, which exposes the status at `error.response.status` (not a top-level `status`), so the previous check silently missed it on cloud — the very path this fix targets. Add a test for the axios error shape. - Move the native-model display prettifier out of the ACP registry constants (it formats OpenHands-kind routing strings, not ACP models) into src/utils/format-model-name.ts as formatNativeModelName(). Fix the trailing-slash edge so it never collapses the chip to an empty string, and cover it with unit tests. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * revert: drop pre-session ACP model-switch band-aid (fixed in SDK #3764) software-agent-sdk#3764 (merged) persists a pre-session ACP model switch at the source: the agent-server route no longer maps to 409, it returns 200, and the first run() uses the switched model. The SDK author notes agent-canvas's "+ New Conversation" pre-run switch "starts working with no change." So the canvas-side 409 handler (try/catch → honest-error toast, the CHAT_INTERFACE$ACP_MODEL_SWITCH_REQUIRES_SESSION string in 15 languages, and the getErrorStatus import) is now redundant — and would be dead/misleading once a release with #3764 is cut. Revert useSwitchAcpModel + its test + the i18n key to main, leaving the live-switch / home-persist behavior the SDK now backs. Issue #1390 finding #2 is resolved upstream. This PR keeps findings #3 (native model chip prettifier) and OpenHands#4 (stale llm_model comment). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Debug Agent <simon@openhands.dev> Co-authored-by: Claude Haiku 4.5 <noreply@anthropic.com>
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 subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
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.
Purpose
Temporary test branch to validate the full Windows GitHub-automation stack before upstream merges.
Includes
pm run dev:test-github-windows-stack, which points the automation backend at jamiechicago312/automation@d20a68c (PR fix: make preset bootstraps shell-neutral across platforms OpenHands/automation#179)
How to test locally
pm ci.
pm run dev:test-github-windows-stack.
Notes
This PR was created by an AI agent (OpenHands) on behalf of the user.