fix(openai-native): address Roomote inline feedback\n\n- Delegate standard GPT-5 SSE event types to shared processor to reduce duplication\n- Add JSDoc for response ID accessors\n- Standardize key error messages for GPT-5 Responses API fallback\n- Extract persistGpt5Metadata() in Task to simplify metadata writes\n- Add malformed JSON SSE parsing test\n

Author: hannesrudolph

src/api/providers/__tests__/openai-native.spec.ts

@@ -1464,4 +1464,54 @@ describe("GPT-5 streaming event coverage (additional)", () => {
 		// @ts-ignore
 		delete global.fetch
 	})
+
+	it("should ignore malformed JSON lines in SSE stream", async () => {
+		const mockFetch = vitest.fn().mockResolvedValue({
+			ok: true,
+			body: new ReadableStream({
+				start(controller) {
+					controller.enqueue(
+						new TextEncoder().encode(
+							'data: {"type":"response.output_item.added","item":{"type":"text","text":"Before"}}\n\n',
+						),
+					)
+					// Malformed JSON line
+					controller.enqueue(
+						new TextEncoder().encode('data: {"type":"response.text.delta","delta":"Bad"\n\n'),
+					)
+					// Valid line after malformed
+					controller.enqueue(
+						new TextEncoder().encode(
+							'data: {"type":"response.output_item.added","item":{"type":"text","text":"After"}}\n\n',
+						),
+					)
+					controller.enqueue(new TextEncoder().encode("data: [DONE]\n\n"))
+					controller.close()
+				},
+			}),
+		})
+		// @ts-ignore
+		global.fetch = mockFetch
+
+		const handler = new OpenAiNativeHandler({
+			apiModelId: "gpt-5-2025-08-07",
+			openAiNativeApiKey: "test-api-key",
+		})
+
+		const systemPrompt = "You are a helpful assistant."
+		const messages: Anthropic.Messages.MessageParam[] = [{ role: "user", content: "Hello!" }]
+		const stream = handler.createMessage(systemPrompt, messages)
+
+		const chunks: any[] = []
+		for await (const chunk of stream) {
+			chunks.push(chunk)
+		}
+
+		// It should not throw and still capture the valid texts around the malformed line
+		const textChunks = chunks.filter((c) => c.type === "text")
+		expect(textChunks.map((c: any) => c.text)).toEqual(["Before", "After"])
+
+		// @ts-ignore
+		delete global.fetch
+	})
 })

src/api/providers/openai-native.ts

@@ -35,6 +35,20 @@ export class OpenAiNativeHandler extends BaseProvider implements SingleCompletio
 	private responseIdPromise: Promise<string | undefined> | undefined
 	private responseIdResolver: ((value: string | undefined) => void) | undefined
 
+	// Event types handled by the shared GPT-5 event processor to avoid duplication
+	private readonly gpt5CoreHandledTypes = new Set<string>([
+		"response.text.delta",
+		"response.output_text.delta",
+		"response.reasoning.delta",
+		"response.reasoning_text.delta",
+		"response.reasoning_summary.delta",
+		"response.reasoning_summary_text.delta",
+		"response.refusal.delta",
+		"response.output_item.added",
+		"response.done",
+		"response.completed",
+	])
+
 	constructor(options: ApiHandlerOptions) {
 		super()
 		this.options = options
@@ -462,7 +476,7 @@ export class OpenAiNativeHandler extends BaseProvider implements SingleCompletio
 					}
 
 					if (!retryResponse.body) {
-						throw new Error("No response body from Responses API retry")
+						throw new Error("GPT-5 Responses API error: No response body from retry request")
 					}
 
 					// Handle the successful retry response
@@ -506,7 +520,7 @@ export class OpenAiNativeHandler extends BaseProvider implements SingleCompletio
 			}
 
 			if (!response.body) {
-				throw new Error("No response body from Responses API")
+				throw new Error("GPT-5 Responses API error: No response body")
 			}
 
 			// Handle streaming response
@@ -603,6 +617,18 @@ export class OpenAiNativeHandler extends BaseProvider implements SingleCompletio
 								this.resolveResponseId(parsed.response.id)
 							}
 
+							// Delegate standard event types to the shared processor to avoid duplication
+							if (parsed?.type && this.gpt5CoreHandledTypes.has(parsed.type)) {
+								for await (const outChunk of this.processGpt5Event(parsed, model)) {
+									// Track whether we've emitted any content so fallback handling can decide appropriately
+									if (outChunk.type === "text" || outChunk.type === "reasoning") {
+										hasContent = true
+									}
+									yield outChunk
+								}
+								continue
+							}
+
 							// Check if this is a complete response (non-streaming format)
 							if (parsed.response && parsed.response.output && Array.isArray(parsed.response.output)) {
 								// Handle complete response in the initial event
@@ -1185,12 +1211,20 @@ export class OpenAiNativeHandler extends BaseProvider implements SingleCompletio
 		return { id: id.startsWith("o3-mini") ? "o3-mini" : id, info, ...params, verbosity: params.verbosity }
 	}
 
-	// Method to get the last response ID for external use
+	/**
+	 * Gets the last GPT-5 response ID captured from the Responses API stream.
+	 * Used for maintaining conversation continuity across requests.
+	 * @returns The response ID, or undefined if not available yet
+	 */
 	getLastResponseId(): string | undefined {
 		return this.lastResponseId
 	}
 
-	// Method to set a response ID for conversation continuity
+	/**
+	 * Sets the last GPT-5 response ID for conversation continuity.
+	 * Typically only used in tests or special flows.
+	 * @param responseId The GPT-5 response ID to store
+	 */
 	setResponseId(responseId: string): void {
 		this.lastResponseId = responseId
 	}

src/core/task/Task.ts

@@ -1751,29 +1751,7 @@ export class Task extends EventEmitter<TaskEvents> implements TaskLike {
 				presentAssistantMessage(this)
 			}
 
-			// Persist GPT‑5 per-turn metadata (response_id, instructions)
-			try {
-				const modelId = this.api.getModel().id
-				if (modelId && modelId.startsWith("gpt-5")) {
-					const lastResponseId: string | undefined = (this.api as any)?.getLastResponseId?.()
-					const idx = findLastIndex(
-						this.clineMessages,
-						(m) => m.type === "say" && (m as any).say === "text" && m.partial !== true,
-					)
-					if (idx !== -1) {
-						const msg = this.clineMessages[idx] as any
-						msg.metadata = msg.metadata ?? {}
-						msg.metadata.gpt5 = {
-							...(msg.metadata.gpt5 ?? {}),
-							response_id: lastResponseId, // This is the response ID generated by THIS turn
-							instructions: this.lastUsedInstructions,
-							reasoning_summary: (reasoningMessage ?? "").trim() || undefined,
-						}
-					}
-				}
-			} catch {
-				// Non-fatal
-			}
+			await this.persistGpt5Metadata(reasoningMessage)
 
 			updateApiReqMsg()
 			await this.saveClineMessages()
@@ -2244,6 +2222,35 @@ export class Task extends EventEmitter<TaskEvents> implements TaskLike {
 		}
 	}
 
+	/**
+	 * Persist GPT-5 per-turn metadata (response_id, instructions, reasoning_summary)
+	 * onto the last complete assistant say("text") message.
+	 */
+	private async persistGpt5Metadata(reasoningMessage?: string): Promise<void> {
+		try {
+			const modelId = this.api.getModel().id
+			if (!modelId || !modelId.startsWith("gpt-5")) return
+
+			const lastResponseId: string | undefined = (this.api as any)?.getLastResponseId?.()
+			const idx = findLastIndex(
+				this.clineMessages,
+				(m) => m.type === "say" && (m as any).say === "text" && m.partial !== true,
+			)
+			if (idx !== -1) {
+				const msg = this.clineMessages[idx] as any
+				msg.metadata = msg.metadata ?? {}
+				msg.metadata.gpt5 = {
+					...(msg.metadata.gpt5 ?? {}),
+					response_id: lastResponseId,
+					instructions: this.lastUsedInstructions,
+					reasoning_summary: (reasoningMessage ?? "").trim() || undefined,
+				}
+			}
+		} catch {
+			// Non-fatal error in metadata persistence
+		}
+	}
+
 	// Getters
 
 	public get cwd() {