fix: preserve audio responses in generated commands

.changeset/audio-binary-output.md

@@ -0,0 +1,12 @@
+---
+"cf": patch
+---
+
+Prepare generated commands to preserve audio API responses
+
+Recognize `audio/*` responses as binary data in the command generator and its
+operation descriptors, including schemas without `format: binary`. Future commands
+generated for audio endpoints will write the exact bytes to stdout so they can be
+redirected to a file. No commands generated from the currently pinned OpenAPI
+specification declare audio responses, so this change does not alter the behavior
+of any currently shipped command.

packages/cli/generator/plugin/emit/index.ts

@@ -182,7 +182,10 @@ export function outputKind(op: OperationInfo): OutputKind {
 		);
 	}
 	if (
-		success.content.some((c) => BINARY.includes(c) || c.startsWith("image/"))
+		success.content.some(
+			(c) =>
+				BINARY.includes(c) || c.startsWith("image/") || c.startsWith("audio/")
+		)
 	) {
 		return "binary";
 	}

packages/cli/generator/plugin/operation.ts

@@ -705,7 +705,8 @@ function wire(method: Method, info: Omit<OperationInfo, "wire">): Wire {
 			ct === "application/octet-stream" ||
 			ct === "application/pdf" ||
 			ct === "application/zip" ||
-			ct.startsWith("image/");
+			ct.startsWith("image/") ||
+			ct.startsWith("audio/");
 		if (binary) {
 			noResponse = false;
 			responseType = "binary-response";

packages/cli/src/__tests__/commands/ai-run.test.ts

@@ -135,12 +135,12 @@ describe("cf ai run", () => {
 		});
 	}
 
-	function rawRunHandler(bytes: Uint8Array) {
+	function rawRunHandler(bytes: Uint8Array, contentType: string) {
 		return http.post(
 			/\/ai\/run$/,
 			() =>
 				new HttpResponse(bytes, {
-					headers: { "content-type": "audio/wav" },
+					headers: { "content-type": contentType },
 				})
 		);
 	}
@@ -168,18 +168,22 @@ describe("cf ai run", () => {
 		expect(stdout()).toContain("hi");
 	});
 
-	it("writes non-JSON model output as exact bytes", async () => {
-		const bytes = Uint8Array.from([82, 73, 70, 70, 0, 255, 128]);
-		const write = vi
-			.spyOn(process.stdout, "write")
-			.mockImplementation(() => true);
-		server.use(schemaHandler(), rawRunHandler(bytes));
+	it.each(["audio/mpeg", "audio/wav", "image/png", "application/octet-stream"])(
+		"writes %s model output as exact bytes",
+		async (contentType) => {
+			const bytes = Uint8Array.from([82, 73, 70, 70, 0, 255, 128]);
+			const write = vi
+				.spyOn(process.stdout, "write")
+				.mockImplementation(() => true);
+			server.use(schemaHandler(), rawRunHandler(bytes, contentType));
 
-		const { exitCode } = await run(MODEL, "--prompt", "hello");
+			const { exitCode } = await run(MODEL, "--prompt", "hello");
 
-		expect(exitCode).toBe(0);
-		expect(write).toHaveBeenCalledWith(Buffer.from(bytes));
-	});
+			expect(exitCode).toBe(0);
+			expect(write).toHaveBeenCalledWith(Buffer.from(bytes));
+			expect(stdout()).toBe("");
+		}
+	);
 
 	it("parses array and object flags as JSON values", async () => {
 		server.use(

packages/cli/src/__tests__/generator/plugin/emit.test.ts

@@ -172,47 +172,50 @@ describe("bodyless raw response emission", () => {
 	);
 });
 
-describe("protobuf response emission", () => {
-	it("writes profile bytes directly while sending the JSON request body", () => {
-		const op = operation(
-			"post",
-			"/accounts/{account_id}/workers/workers/{worker_id}/versions/{version_id}/profile",
-			{
-				pathParams: [
-					param("account_id"),
-					param("worker_id"),
-					param("version_id"),
-				],
-				bodyParams: [
-					bodyParam("duration_ms", "number", ["duration_ms"], {
-						required: true,
-					}),
-				],
-				hasRequestBody: true,
-				requestContentTypes: ["application/json"],
-				requestBodyRequired: ["duration_ms"],
-				responses: [["200", { content: ["application/vnd.google.protobuf"] }]],
-			}
-		);
-		const ctx = context(
-			method("profile", "workers.versions.profile"),
-			"workers",
-			op
-		);
-		expect(ctx.outputKind).toBe("binary");
-		const output = handler(ctx).join("\n");
-		expect(output).toContain("fetchRawBytes(");
-		expect(output).toContain("method: 'POST'");
-		expect(output).toContain("body: JSON.stringify(bodyData)");
-		expect(output).toContain(
-			"body: Object.keys(bodyData).length > 0 ? bodyData : undefined"
-		);
-		expect(output).toContain("contentType: 'application/json'");
-		expect(output).toContain(
-			"writeRawOutput(argv.text === true ? __cfRawBytes.toString('utf-8') : __cfRawBytes)"
-		);
-		expect(output).not.toContain("formatOutput(");
-	});
+describe("binary response emission", () => {
+	it.each(["application/vnd.google.protobuf", "audio/mpeg", "audio/wav"])(
+		"writes %s bytes directly while sending the JSON request body",
+		(contentType) => {
+			const op = operation(
+				"post",
+				"/accounts/{account_id}/workers/workers/{worker_id}/versions/{version_id}/profile",
+				{
+					pathParams: [
+						param("account_id"),
+						param("worker_id"),
+						param("version_id"),
+					],
+					bodyParams: [
+						bodyParam("duration_ms", "number", ["duration_ms"], {
+							required: true,
+						}),
+					],
+					hasRequestBody: true,
+					requestContentTypes: ["application/json"],
+					requestBodyRequired: ["duration_ms"],
+					responses: [["200", { content: [contentType] }]],
+				}
+			);
+			const ctx = context(
+				method("profile", "workers.versions.profile"),
+				"workers",
+				op
+			);
+			expect(ctx.outputKind).toBe("binary");
+			const output = handler(ctx).join("\n");
+			expect(output).toContain("fetchRawBytes(");
+			expect(output).toContain("method: 'POST'");
+			expect(output).toContain("body: JSON.stringify(bodyData)");
+			expect(output).toContain(
+				"body: Object.keys(bodyData).length > 0 ? bodyData : undefined"
+			);
+			expect(output).toContain("contentType: 'application/json'");
+			expect(output).toContain(
+				"writeRawOutput(argv.text === true ? __cfRawBytes.toString('utf-8') : __cfRawBytes)"
+			);
+			expect(output).not.toContain("formatOutput(");
+		}
+	);
 });
 
 describe("variant prompts", () => {

packages/cli/src/__tests__/generator/plugin/forge-model.test.ts

@@ -393,6 +393,47 @@ describe("the forge model", () => {
 		expect(code).toContain(".conflicts('id', ['name'])");
 	});
 
+	it.each(["audio/mpeg", "audio/wav"])(
+		"preserves %s responses without a binary schema format",
+		(contentType) => {
+			const files = generateFiles(
+				doc({
+					"/speech": {
+						post: {
+							"x-fern-sdk-group-name": "speech",
+							"x-fern-sdk-method-name": "create",
+							summary: "Generate speech",
+							requestBody: {
+								content: {
+									"application/json": {
+										schema: {
+											type: "object",
+											properties: { prompt: { type: "string" } },
+										},
+									},
+								},
+							},
+							responses: {
+								"200": {
+									description: "Audio bytes",
+									content: { [contentType]: { schema: { type: "string" } } },
+								},
+							},
+						},
+					},
+				})
+			);
+			const command = files.get("speech/create.ts") ?? "";
+			expect(command).toContain("fetchRawBytes(");
+			expect(command).toContain("method: 'POST'");
+			expect(command).toContain("writeRawOutput(");
+			expect(command).not.toContain("formatOutput(");
+			expect(operation(files, "speech/create")).toContain(
+				'response: "binary-response"'
+			);
+		}
+	);
+
 	it("describes the operation a typed call sends", () => {
 		const textGet = doc({
 			"/accounts/{account_id}/things/{thing_id}": {

packages/cli/src/__tests__/generator/plugin/output-kind.test.ts

@@ -8,6 +8,9 @@ describe("outputKind", () => {
 		["application/octet-stream", "binary"],
 		["application/zip", "binary"],
 		["image/png", "binary"],
+		["audio/mpeg", "binary"],
+		["audio/wav", "binary"],
+		["audio/ogg", "binary"],
 		["text/plain", "text"],
 		["text/csv", "text"],
 		["application/json", "json"],