maoda-commander-tt 0.0.41 → 0.0.43

package/README.md

@@ -61,7 +61,7 @@ pnpm start -- --help
 ## AI 音频生成
 
 `ai audio` 接受一段自然语言提示词并等待 XYQ 返回 MP3。参考音频可选；提供时，
-提示词可以用“参考音频1”描述要复用的音色。
+提示词需自行用 `@音频1` 引用它，命令不会改写提示词。
 
 ```bash
 # 纯文本生成音频
@@ -69,11 +69,11 @@ tt ai audio '萝莉音说：“我今天不回来了”'
 
 # 参考一段本地音频的音色
 tt ai audio --audio ./reference.mp3 \
-  '参考音频1的音色，说：“我生气了，哼”'
+  '@音频1 参考它的音色，说：“我生气了，哼”'
 
 # 等待生成并下载到本地；省略 --output 时返回远端地址
 tt ai audio --audio ./reference.mp3 --output ./result.mp3 \
-  '参考音频1的音色，说：“我生气了，哼”'
+  '@音频1 参考它的音色，说：“我生气了，哼”'
 ```
 
 命令固定使用 `seedaudio_1.0`、MP3 和 44.1 kHz，最多等待 60 秒。参考音频直接

package/dist/main.js

@@ -3289,6 +3289,7 @@ var ASSET_TYPE = {
   image: "2",
   audio: "4"
 };
+var VIDEO_POLL_INTERVAL_SECONDS = 3;
 var IMAGE_EXTENSIONS2 = /* @__PURE__ */ new Set([
   ".jpg",
   ".jpeg",
@@ -3353,7 +3354,7 @@ var XyqSession = class {
     const submittedAt = this._now();
     const body = buildSubmitAudioBody({
       prompt,
-      pippitAssetId: normalizeString2(input.pippitAssetId),
+      referenceAudioPippitAssetId: normalizeString2(input.pippitAssetId),
       threadId,
       clientRunId,
       messageId,
@@ -3584,15 +3585,11 @@ var XyqSession = class {
     ) : void 0) ?? runList.at(-1);
     const runRecord = isRecord2(run) ? run : {};
     const serverRunId = normalizeString2(runRecord.run_id);
-    const queueInfo = await this._queryQueueInfoBestEffort(
-      threadId,
-      serverRunId ?? requestedRunId
-    );
     const entryList = Array.isArray(runRecord.entry_list) ? runRecord.entry_list : [];
     const artifacts = entryList.map(extractVideoArtifact).filter((artifact) => artifact !== void 0);
     const providerState = mapXyqRunState(toNumber(runRecord.state));
     const state = hasUsableVideoOutput(artifacts) ? "completed" : providerState;
-    const retryAfterSeconds = isTerminalXyqTaskState(state) ? 0 : resolveRetryAfterSeconds(queueInfo ?? runRecord.RunQueueInfo);
+    const retryAfterSeconds = isTerminalXyqTaskState(state) ? 0 : VIDEO_POLL_INTERVAL_SECONDS;
     return ok19({
       taskId: createXyqTaskId(threadId),
       threadId,
@@ -3600,7 +3597,7 @@ var XyqSession = class {
       state,
       retryAfterSeconds,
       artifacts,
-      ...queueInfo !== void 0 ? { queueInfo: camelizeValue(queueInfo) } : runRecord.RunQueueInfo !== void 0 ? { queueInfo: camelizeValue(runRecord.RunQueueInfo) } : {},
+      ...runRecord.RunQueueInfo !== void 0 ? { queueInfo: camelizeValue(runRecord.RunQueueInfo) } : {},
       ...runRecord.fail_reason !== void 0 ? { failReason: camelizeValue(runRecord.fail_reason) } : {},
       raw: camelizeValue(threadResult.value.data)
     });
@@ -3815,31 +3812,6 @@ var XyqSession = class {
     } : {};
     return ok19(headers);
   }
-  async _queryQueueInfoBestEffort(threadId, runId) {
-    const result = await requestXyqJson(
-      this._fetcher,
-      `${XYQ_ORIGIN}/api/biz/v1/agent/get_user_queue_info`,
-      {
-        method: "POST",
-        headers: {
-          ...this._headers(),
-          "content-type": "application/json"
-        },
-        body: JSON.stringify({
-          uid: this.context.uid,
-          thread_id: threadId,
-          ...runId ? { run_id: runId } : {},
-          need_earliest_submitted_working_run: true
-        })
-      },
-      {
-        requestFailedCode: XyqErrorCode.VideoQueryRequestFailed,
-        responseInvalidCode: XyqErrorCode.VideoQueryResponseInvalid,
-        operation: "\u67E5\u8BE2 XYQ \u89C6\u9891\u4EFB\u52A1\u961F\u5217"
-      }
-    );
-    return result.ok ? result.value.data : void 0;
-  }
 };
 function buildSubmitAudioBody(input) {
   const contentData = {
@@ -3853,7 +3825,14 @@ function buildSubmitAudioBody(input) {
       loudness_rate: 0,
       pitch_rate: 0
     },
-    ...input.pippitAssetId ? { pippit_asset_id: input.pippitAssetId } : {}
+    ...input.referenceAudioPippitAssetId ? {
+      references: [
+        {
+          type: "audio",
+          pippit_asset_id: input.referenceAudioPippitAssetId
+        }
+      ]
+    } : {}
   };
   const runExtra = {
     babi_param: {
@@ -4098,19 +4077,6 @@ function extractVideoArtifact(entry) {
     raw: camelizeValue(artifact)
   };
 }
-function resolveRetryAfterSeconds(queueInfo) {
-  const parsed = parseJsonRecord(queueInfo) ?? queueInfo;
-  const value = deepFindStringByKey(parsed, [
-    "estimated_wait_seconds",
-    "wait_seconds",
-    "queue_wait_seconds"
-  ]);
-  const seconds = Number(value ?? 0);
-  if (Number.isFinite(seconds) && seconds > 0) {
-    return Math.max(1, Math.min(Math.ceil(seconds), 5));
-  }
-  return 3;
-}
 function deepCollectAudioOutputs(value) {
   const results = /* @__PURE__ */ new Map();
   collectAudioOutputs(value, void 0, results, 0);
@@ -4290,36 +4256,6 @@ function collectExplicitVideo(value, inheritedPippitAssetId, results, depth) {
 function videoQualityScore(output) {
   return (output.width ?? 0) * (output.height ?? 0);
 }
-function deepFindStringByKey(value, keys, depth = 0) {
-  if (depth > 6) {
-    return void 0;
-  }
-  if (Array.isArray(value)) {
-    for (const item of value) {
-      const found = deepFindStringByKey(item, keys, depth + 1);
-      if (found) {
-        return found;
-      }
-    }
-    return void 0;
-  }
-  if (!isRecord2(value)) {
-    return void 0;
-  }
-  for (const key of keys) {
-    const found = normalizeString2(value[key]);
-    if (found) {
-      return found;
-    }
-  }
-  for (const child of Object.values(value)) {
-    const found = deepFindStringByKey(child, keys, depth + 1);
-    if (found) {
-      return found;
-    }
-  }
-  return void 0;
-}
 function parseJsonRecord(value) {
   if (isRecord2(value)) {
     return value;
@@ -4757,7 +4693,7 @@ var AiAudioCommand = class extends AiBaseCommand {
   }
   _configureOptions(cmd) {
     cmd.addOption(
-      new Option8("--audio <path>", "\u53EF\u9009\u53C2\u8003\u97F3\u9891\uFF0C\u6700\u591A\u4E00\u4E2A").argParser(appendAudioPath)
+      new Option8("--audio <path>", "\u53EF\u9009\u53C2\u8003\u97F3\u9891\uFF0C\u6700\u591A\u4E00\u4E2A\uFF1B\u63D0\u793A\u8BCD\u9700\u7528 @\u97F3\u98911 \u5F15\u7528").argParser(appendAudioPath)
     );
     cmd.addOption(
       new Option8(
@@ -4793,7 +4729,7 @@ var AiAudioCommand = class extends AiBaseCommand {
         return this._fail(sessionResult.code, sessionResult.msg);
       }
       const session = sessionResult.value;
-      let pippitAssetId;
+      let referenceAudioPippitAssetId;
       if (audioPath) {
         const uploadResult = await session.uploadAudio(audioPath, {
           useCache: false
@@ -4801,14 +4737,14 @@ var AiAudioCommand = class extends AiBaseCommand {
         if (!uploadResult.ok) {
           return this._fail(uploadResult.code, uploadResult.msg);
         }
-        pippitAssetId = uploadResult.value.pippitAssetId?.trim();
-        if (!pippitAssetId) {
+        referenceAudioPippitAssetId = uploadResult.value.pippitAssetId?.trim();
+        if (!referenceAudioPippitAssetId) {
           return this._fail(1, "\u53C2\u8003\u97F3\u9891\u4E0A\u4F20\u6210\u529F\uFF0C\u4F46\u672A\u8FD4\u56DE pippitAssetId");
         }
       }
       const submitResult = await session.submitAudio({
         prompt,
-        ...pippitAssetId ? { pippitAssetId } : {}
+        ...referenceAudioPippitAssetId ? { pippitAssetId: referenceAudioPippitAssetId } : {}
       });
       if (!submitResult.ok) {
         return this._fail(submitResult.code, submitResult.msg);