fix: Azure GPT-5 responses replay ordering

packages/opencode/src/provider/provider.ts

@@ -1428,8 +1428,13 @@ export namespace Provider {
             const combined = signals.length === 0 ? null : signals.length === 1 ? signals[0] : AbortSignal.any(signals)
             if (combined) opts.signal = combined
 
-            // Strip openai itemId metadata following what codex does
-            if (model.api.npm === "@ai-sdk/openai" && opts.body && opts.method === "POST") {
+            // Strip responses item ids unless we are deliberately replaying stored Azure items.
+            // Azure uses a dedicated SDK package, so the check must cover both providers.
+            if (
+              (model.api.npm === "@ai-sdk/openai" || model.api.npm === "@ai-sdk/azure") &&
+              opts.body &&
+              opts.method === "POST"
+            ) {
               const body = JSON.parse(opts.body as string)
               const isAzure = model.providerID.includes("azure")
               const keepIds = isAzure && body.store === true

packages/opencode/src/provider/sdk/copilot/responses/convert-to-openai-responses-input.ts

@@ -6,7 +6,7 @@ import {
 } from "@ai-sdk/provider"
 import { convertToBase64, parseProviderOptions } from "@ai-sdk/provider-utils"
 import { z } from "zod/v4"
-import type { OpenAIResponsesInput, OpenAIResponsesReasoning } from "./openai-responses-api-types"
+import type { OpenAIResponsesInput, OpenAIResponsesInputItem, OpenAIResponsesReasoning } from "./openai-responses-api-types"
 import { localShellInputSchema, localShellOutputSchema } from "./tool/local-shell"
 
 /**
@@ -118,13 +118,58 @@ export async function convertToOpenAIResponsesInput({
       }
 
       case "assistant": {
+        const assistantInput: OpenAIResponsesInput = []
         const reasoningMessages: Record<string, OpenAIResponsesReasoning> = {}
         const toolCallParts: Record<string, LanguageModelV3ToolCallPart> = {}
+        let pendingOutput: OpenAIResponsesInputItem | undefined
+        let pendingReasoning: OpenAIResponsesInputItem | undefined
+
+        const pushAssistantOutput = (item: OpenAIResponsesInputItem) => {
+          if (pendingReasoning !== undefined) {
+            assistantInput.push(pendingReasoning)
+            pendingReasoning = undefined
+            assistantInput.push(item)
+            return
+          }
+
+          if (pendingOutput !== undefined) {
+            assistantInput.push(pendingOutput)
+          }
+
+          pendingOutput = item
+        }
+
+        const pushAssistantReasoning = (item: OpenAIResponsesInputItem) => {
+          if (pendingOutput !== undefined) {
+            assistantInput.push(item)
+            assistantInput.push(pendingOutput)
+            pendingOutput = undefined
+            return
+          }
+
+          if (pendingReasoning !== undefined) {
+            assistantInput.push(pendingReasoning)
+          }
+
+          pendingReasoning = item
+        }
+
+        const flushAssistantPending = () => {
+          if (pendingReasoning !== undefined) {
+            assistantInput.push(pendingReasoning)
+            pendingReasoning = undefined
+          }
+
+          if (pendingOutput !== undefined) {
+            assistantInput.push(pendingOutput)
+            pendingOutput = undefined
+          }
+        }
 
         for (const part of content) {
           switch (part.type) {
             case "text": {
-              input.push({
+              pushAssistantOutput({
                 role: "assistant",
                 content: [{ type: "output_text", text: part.text }],
                 id: (part.providerOptions?.openai?.itemId as string) ?? undefined,
@@ -140,7 +185,7 @@ export async function convertToOpenAIResponsesInput({
 
               if (hasLocalShellTool && part.toolName === "local_shell") {
                 const parsedInput = localShellInputSchema.parse(part.input)
-                input.push({
+                pushAssistantOutput({
                   type: "local_shell_call",
                   call_id: part.toolCallId,
                   id: (part.providerOptions?.openai?.itemId as string) ?? undefined,
@@ -157,7 +202,7 @@ export async function convertToOpenAIResponsesInput({
                 break
               }
 
-              input.push({
+              pushAssistantOutput({
                 type: "function_call",
                 call_id: part.toolCallId,
                 name: part.toolName,
@@ -171,7 +216,7 @@ export async function convertToOpenAIResponsesInput({
             case "tool-result": {
               if (store) {
                 // use item references to refer to tool results from built-in tools
-                input.push({ type: "item_reference", id: part.toolCallId })
+                pushAssistantOutput({ type: "item_reference", id: part.toolCallId })
               } else {
                 warnings.push({
                   type: "other",
@@ -197,7 +242,7 @@ export async function convertToOpenAIResponsesInput({
                 if (store) {
                   if (reasoningMessage === undefined) {
                     // use item references to refer to reasoning (single reference)
-                    input.push({ type: "item_reference", id: reasoningId })
+                    pushAssistantReasoning({ type: "item_reference", id: reasoningId })
 
                     // store unused reasoning message to mark id as used
                     reasoningMessages[reasoningId] = {
@@ -231,7 +276,7 @@ export async function convertToOpenAIResponsesInput({
                       encrypted_content: providerOptions?.reasoningEncryptedContent,
                       summary: summaryParts,
                     }
-                    input.push(reasoningMessages[reasoningId])
+                    pushAssistantReasoning(reasoningMessages[reasoningId])
                   } else {
                     reasoningMessage.summary.push(...summaryParts)
                   }
@@ -247,6 +292,9 @@ export async function convertToOpenAIResponsesInput({
           }
         }
 
+        flushAssistantPending()
+        input.push(...assistantInput)
+
         break
       }
 

packages/opencode/test/provider/openai-responses-input.test.ts

@@ -0,0 +1,103 @@
+import { describe, expect, test } from "bun:test"
+import { convertToOpenAIResponsesInput } from "../../src/provider/sdk/copilot/responses/convert-to-openai-responses-input"
+
+describe("convertToOpenAIResponsesInput", () => {
+  test("emits reasoning before an assistant message when replay order arrives reversed", async () => {
+    const result = await convertToOpenAIResponsesInput({
+      prompt: [
+        {
+          role: "assistant",
+          content: [
+            {
+              type: "text",
+              text: "Hello",
+              providerOptions: {
+                openai: {
+                  itemId: "msg_123",
+                },
+              },
+            },
+            {
+              type: "reasoning",
+              text: "thinking",
+              providerOptions: {
+                copilot: {
+                  itemId: "rs_123",
+                  reasoningEncryptedContent: "encrypted",
+                },
+              },
+            },
+          ],
+        },
+      ] as any,
+      systemMessageMode: "system",
+      store: false,
+    })
+
+    expect(result.input).toMatchObject([
+      {
+        type: "reasoning",
+        id: "rs_123",
+        encrypted_content: "encrypted",
+        summary: [{ type: "summary_text", text: "thinking" }],
+      },
+      {
+        role: "assistant",
+        id: "msg_123",
+        content: [{ type: "output_text", text: "Hello" }],
+      },
+    ])
+  })
+
+  test("emits reasoning before a function call when replay order arrives reversed", async () => {
+    const result = await convertToOpenAIResponsesInput({
+      prompt: [
+        {
+          role: "assistant",
+          content: [
+            {
+              type: "tool-call",
+              toolCallId: "call_123",
+              toolName: "read_file",
+              input: { filePath: "/README.md" },
+              providerExecuted: false,
+              providerOptions: {
+                openai: {
+                  itemId: "fc_123",
+                },
+              },
+            },
+            {
+              type: "reasoning",
+              text: "thinking",
+              providerOptions: {
+                copilot: {
+                  itemId: "rs_456",
+                  reasoningEncryptedContent: "encrypted",
+                },
+              },
+            },
+          ],
+        },
+      ] as any,
+      systemMessageMode: "system",
+      store: false,
+    })
+
+    expect(result.input).toMatchObject([
+      {
+        type: "reasoning",
+        id: "rs_456",
+        encrypted_content: "encrypted",
+        summary: [{ type: "summary_text", text: "thinking" }],
+      },
+      {
+        type: "function_call",
+        id: "fc_123",
+        call_id: "call_123",
+        name: "read_file",
+        arguments: '{"filePath":"/README.md"}',
+      },
+    ])
+  })
+})