Conversation
ammario
enabled auto-merge (squash)
October 7, 2025 02:20
OpenAI Responses API was failing with error: 'Item rs_* of type reasoning was provided without its required following item' This occurred when using reasoning models (gpt-5, o3, o4-mini) with tool calls in multi-turn conversations. Root cause: When using reasoning models with tool calls, OpenAI requires including 'reasoning.encrypted_content' in the include option to properly preserve reasoning context across conversation steps. Solution: Add include: ['reasoning.encrypted_content'] to OpenAI provider options when reasoningEffort is configured. Refs: https://sdk.vercel.ai/providers/ai-sdk-providers/openai#responses-models
ammario
added a commit
that referenced
this pull request
Oct 7, 2025
## Problem After merging PR #59, the OpenAI reasoning error still occurred: ``` Item 'rs_01e84f5b161f6bce0068e480e7778481a3a2c3b6234d6bb7c6' of type 'reasoning' was provided without its required following item. ``` ## Root Cause Analysis **The issue was OpenAI-specific.** Anthropic reasoning and OpenAI reasoning work differently: ### Anthropic - Uses **text-based reasoning parts** in message content - These **SHOULD be sent back** to the API (via `sendReasoning: true` option, which defaults to true) - The model uses historical reasoning context to improve responses ### OpenAI - Uses **encrypted reasoning items** with IDs (e.g., `rs_*`) - These are managed automatically via `previous_response_id` - Sending Anthropic-style text reasoning parts creates **orphaned reasoning items** that cause API errors - Per OpenAI docs: "In typical multi-turn conversations, you don't need to include reasoning items or tokens—the model is trained to produce the best output without them" ## Solution **Strip reasoning parts ONLY for OpenAI**, before converting CmuxMessages to ModelMessages. - Added `stripReasoningForOpenAI()` function for OpenAI-specific processing - Apply it conditionally based on provider in `aiService.ts` - Keeps Anthropic behavior intact (reasoning sent via `sendReasoning`) ## Changes 1. **New function**: `stripReasoningForOpenAI()` in `modelMessageTransform.ts` 2. **Updated `aiService.ts`**: Apply reasoning stripping only for OpenAI provider 3. **Reverted previous change**: `filterEmptyAssistantMessages()` no longer strips all reasoning 4. **Added detailed comments** explaining provider-specific differences This ensures: - ✅ OpenAI doesn't get orphaned reasoning items - ✅ Anthropic still receives reasoning context - ✅ Provider-specific behavior is clearly documented
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.
Problem
OpenAI Responses API was intermittently failing with error:
This occurred when using reasoning models (gpt-5, o3, o4-mini) with tool calls in multi-turn conversations.
Root Cause
From the Vercel AI SDK documentation:
Even though we're using the default
store: trueandpreviousResponseIdfor persistence, we still need to explicitly include reasoning encrypted content when tool calls are involved. The reasoning items have IDs (likers_*) that must be properly linked to their following items.Solution
Added
include: ['reasoning.encrypted_content']to OpenAI provider options whenreasoningEffortis configured (meaning reasoning is enabled).This ensures reasoning context is properly preserved across multi-turn conversations with tool calls.
Changes
buildProviderOptions()insrc/utils/ai/providerOptions.tsincludewhen reasoning is actually enabledTesting
Not adding automated tests as the error was intermittent and difficult to reliably reproduce. The fix is based directly on OpenAI/Vercel SDK documentation requirements.