fix(ai): skip reasoning-only responses replay #838
Closed
Conversation
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
This was referenced Jan 19, 2026
Open
Owner
|
Thanks @steipete for the report and investigation. Fixed this differently: we now skip errored/aborted assistant messages entirely in transform-messages.ts rather than filtering at the provider level. This way the fix covers openai-responses, openai-codex-responses, and any future providers. Also removed the strictResponsesPairing workaround from #768 since it's no longer needed. |
badlogic
added a commit
that referenced
this pull request
Jan 19, 2026
Fixes OpenAI Responses 400 error 'reasoning without following item' by skipping errored/aborted assistant messages entirely rather than filtering at the provider level. This covers openai-responses, openai-codex-responses, and future providers. Removes strictResponsesPairing compat option (no longer needed). Closes #838
badlogic
added a commit
that referenced
this pull request
Jan 19, 2026
barnierg76
added a commit
to barnierg76/pi-mono
that referenced
this pull request
Jan 25, 2026
…sistant messages When an assistant message with tool_use blocks is skipped due to error or aborted stopReason, the corresponding tool_result messages were not being filtered out, causing Anthropic API rejections: "unexpected tool_use_id found in tool_result blocks". This fix: - Tracks tool IDs from skipped assistant messages in a Set - Filters out tool results that reference these skipped tool IDs - Ensures strict Anthropic API requirement: every tool_result must have a corresponding tool_use in a preceding assistant message Fixes issue introduced in v0.49.2 (issue badlogic#838) where errored messages were skipped to fix OpenAI compatibility, but orphaned tool results remained. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
5 tasks
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
I have two reports where we corrupted the session for OpenAI:
OpenAI Responses 400: “Item ‘rs_…’ of type ‘reasoning’ was provided without its required following item.”
Trigger: history replay includes assistant turns with only thinking/reasoning (often from aborted runs). pi-ai replays thinking Signature as a reasoning item even when the same turn has no message or function_call.
OpenAI Responses reasoning item replay causes 400 (missing following item) moltbot/moltbot#1126
OpenAI GPT-5.2: reasoning item without required following item moltbot/moltbot#1158
I did check and found no docs (ofc) but forum entries and it happening twice seems this is a real issue.
Don't have session logs.
vibed pr. fix is small and for now I ship a local workaround.