import { z } from "zod";
import { generateSecMsGecToken_V2, SEC_MS_GEC_VERSION_V2, TRUSTED_CLIENT_TOKEN } from "node-edge-tts/dist/drm";
import { speechVoiceSchema, type SpeechVoice } from "../src/shared/tts/voices";

const upstreamVoicesSchema = z.array(z.object({
  ShortName: z.string().min(1).max(120),
  FriendlyName: z.string().min(1).max(200),
  Locale: z.string().min(2).max(40),
  Gender: z.enum(["Female", "Male"])
})).min(1).max(2_000);

export async function fetchEdgeTtsVoices(): Promise<SpeechVoice[]> {
  const url = new URL("https://speech.platform.bing.com/consumer/speech/synthesize/readaloud/voices/list");
  url.searchParams.set("trustedclienttoken", TRUSTED_CLIENT_TOKEN);
  url.searchParams.set("Sec-MS-GEC", generateSecMsGecToken_V2());
  url.searchParams.set("Sec-MS-GEC-Version", SEC_MS_GEC_VERSION_V2);
  const response = await fetch(url, { signal: AbortSignal.timeout(10_000) });
  if (!response.ok) throw new Error(`Edge TTS voice catalog returned ${response.status}`);
  const voices = upstreamVoicesSchema.parse(await response.json()).filter((voice) => voice.Locale === "zh-CN");
  if (!voices.length) throw new Error("Edge TTS returned no Mandarin voices");
  return voices.map((voice) => speechVoiceSchema.parse({
    id: voice.ShortName, name: voice.FriendlyName, locale: voice.Locale, gender: voice.Gender
  }));
}

export function createEdgeTtsVoiceCatalog(
  load: () => Promise<SpeechVoice[]> = fetchEdgeTtsVoices,
  now: () => number = Date.now
) {
  let cached: SpeechVoice[] | undefined;
  let expiresAt = 0;
  let pending: Promise<SpeechVoice[]> | undefined;
  return () => {
    if (cached && now() < expiresAt) return Promise.resolve(cached);
    if (!pending) {
      pending = Promise.resolve().then(load).then((voices) => {
        cached = voices;
        expiresAt = now() + 60 * 60 * 1_000;
        return voices;
      }).finally(() => { pending = undefined; });
    }
    return pending;
  };
}