feat(web): generateDialog client with per-speaker multipart

web/src/lib/api.ts

@@ -95,3 +95,43 @@ export function streamActiveEvents(
   };
   return () => es.close();
 }
+
+export type DialogSpeakerInput = {
+  letter: "A" | "B" | "C" | "D";
+  reference: Blob;
+};
+
+export type DialogInput = {
+  engineId: string;
+  text: string;
+  language?: string;
+  params: Record<string, unknown>;
+  speakers: DialogSpeakerInput[];
+};
+
+export type DialogResult = {
+  blob: Blob;
+  seedUsed: number | null;
+};
+
+export async function generateDialog(input: DialogInput): Promise<DialogResult> {
+  const fd = new FormData();
+  fd.set("text", input.text);
+  fd.set("engine_id", input.engineId);
+  fd.set("params", JSON.stringify(input.params ?? {}));
+  if (input.language) fd.set("language", input.language);
+  for (const s of input.speakers) {
+    fd.set(`reference_wav_${s.letter.toLowerCase()}`, s.reference, `${s.letter}.wav`);
+  }
+  const r = await fetch("/api/generate/dialog", { method: "POST", body: fd });
+  if (!r.ok) {
+    const err = await r.json().catch(() => ({}));
+    const code = err?.error?.code ?? `dialog: ${r.status}`;
+    const msg = err?.error?.message;
+    throw new Error(msg ? `${code}: ${msg}` : code);
+  }
+  const seedHeader = r.headers.get("x-seed-used");
+  const seedUsed = seedHeader != null ? Number(seedHeader) : null;
+  const blob = await r.blob();
+  return { blob, seedUsed };
+}

web/src/test/api.test.ts

@@ -1,5 +1,6 @@
 import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
 import { activateModel, generate, getActiveModel, listModels } from "@/lib/api";
+import { generateDialog } from "@/lib/api";
 
 const fetchMock = vi.fn();
 
@@ -65,3 +66,45 @@ describe("api", () => {
     ).rejects.toThrow(/model_not_found/);
   });
 });
+
+describe("generateDialog", () => {
+  it("posts multipart with engine_id and per-speaker clips", async () => {
+    fetchMock.mockResolvedValue(
+      new Response("RIFFOK", {
+        status: 200,
+        headers: { "X-Seed-Used": "33" },
+      }),
+    );
+    const out = await generateDialog({
+      engineId: "x",
+      text: "SPEAKER A: hi\nSPEAKER B: hi",
+      params: { temperature: 0.8 },
+      speakers: [
+        { letter: "A", reference: new Blob(["a"], { type: "audio/wav" }) },
+        { letter: "B", reference: new Blob(["b"], { type: "audio/wav" }) },
+      ],
+    });
+    expect(out.seedUsed).toBe(33);
+    expect(typeof out.blob.size).toBe("number");
+    const call = fetchMock.mock.calls[0];
+    expect(call[0]).toBe("/api/generate/dialog");
+    const body = call[1].body as FormData;
+    expect(body.get("engine_id")).toBe("x");
+    expect(body.get("text")).toContain("SPEAKER A:");
+    expect(body.get("reference_wav_a")).toBeInstanceOf(Blob);
+    expect(body.get("reference_wav_b")).toBeInstanceOf(Blob);
+  });
+
+  it("forwards language only when provided", async () => {
+    fetchMock.mockResolvedValue(new Response("RIFF", { status: 200 }));
+    await generateDialog({
+      engineId: "x",
+      text: "SPEAKER A: hi",
+      language: "fr",
+      params: {},
+      speakers: [{ letter: "A", reference: new Blob(["a"]) }],
+    });
+    const body = fetchMock.mock.calls[0][1].body as FormData;
+    expect(body.get("language")).toBe("fr");
+  });
+});