Fix Claude Code follow-up/compact 400s and duplicated assistant text - #1
Open
zhaohui-yang wants to merge 1 commit into
Open
Fix Claude Code follow-up/compact 400s and duplicated assistant text#1zhaohui-yang wants to merge 1 commit into
zhaohui-yang wants to merge 1 commit into
Conversation
Ignore historical tool_result when deciding resume, treat stale ids as a new run instead of 400, and do not append result.result after streamed assistant text.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Claude Code (and other Anthropic clients) resend the full message list every turn, including
tool_resultblocks from already-finished rounds. Three bugs in the protocol adapter/tool relay show up on that path:Duplicated assistant text. Streaming already forwarded
assistantdeltas, thenrun.wait().result.resultwas appended again because#hasTextlooked atsink.parts. The Anthropic SSE writer stores text onaccum, so the check was always false after a streamed turn.Follow-up 400
no matching pending tool call.digToolResultsscanned the entiremessagesarray. Historicaltool_resultids no longer map to a liveRelayTurn.Autocompact then fails with the same 400. Compact requests often end with finished
tool_resultblocks after the last assistant message. Returning 400 aborts compact and Claude Code surfacesPrompt is too long · automatic compaction failed. Stale ids now log a warning and start a new run (folded prompt) instead of 400.Tradeoff on (3): a client that retries tool results for a turn that already expired will bill a new Cloud Agents run rather than get 400. That is the interoperable behavior for Claude Code; tests updated accordingly.
Reproduction:
/v1/messages, run a turn that calls a client tool, then send a second user prompt.