@lingjingai/scriptctl 0.63.9 → 0.63.10

package/changes/0.63.10.md

@@ -0,0 +1,4 @@
+# 0.63.10
+
+- 原片没有音轨时记录证据不足，保留台词与时间并继续画面审核，不再尝试无效音频提取。
+- 审核视频及结构化请求使用服务端默认输出额度。

package/dist/review-agent.js

@@ -2474,12 +2474,7 @@ async function generateReviewJson(request) {
         { text: request.prompt }
       ] }],
       ...request.system ? { systemInstruction: { parts: [{ text: request.system }] } } : {},
-      generationConfig: {
-        temperature: 0,
-        responseMimeType: "application/json",
-        responseJsonSchema: schema,
-        ...request.maxOutputTokens ? { maxOutputTokens: request.maxOutputTokens } : {}
-      }
+      generationConfig: { temperature: 0, responseMimeType: "application/json", responseJsonSchema: schema }
     } : {
       model: route.model,
       input: [{ role: "user", content: [
@@ -2487,7 +2482,6 @@ async function generateReviewJson(request) {
         { type: "input_text", text: request.prompt }
       ] }],
       ...request.system ? { instructions: request.system } : {},
-      ...request.maxOutputTokens ? { max_output_tokens: request.maxOutputTokens } : {},
       text: { format: { type: "json_schema", name: "review", schema } }
     };
     for (let attempt = 1; attempt <= route.maxAttempts; attempt++) {
@@ -2579,7 +2573,6 @@ async function watchWindow(video, from, to, question, opts) {
       video: { data: fs3.readFileSync(clip2.path, "base64"), fps: CLIP_FPS },
       schema: z3.object({ answer: z3.string().min(1) }),
       timeoutMs: 18e4,
-      maxOutputTokens: 2048,
       callKey: opts.callKey
     });
     return { answer: result.value.answer, usage: result.usage.map((row) => ({
@@ -2613,6 +2606,20 @@ var wordsSchema = z4.array(z4.object({
 })).transform((words) => words.filter((word) => word.text.trim().length > 0)).pipe(z4.array(timedTextSchema));
 var utteranceSchema = timedTextSchema.safeExtend({ words: wordsSchema.optional() });
 var responseSchema = z4.object({ result: z4.object({ utterances: z4.array(utteranceSchema) }) });
+async function hasAudioTrack(video) {
+  const { stdout } = await exec("ffprobe", [
+    "-v",
+    "error",
+    "-select_streams",
+    "a:0",
+    "-show_entries",
+    "stream=index",
+    "-of",
+    "csv=p=0",
+    video
+  ], { timeout: 3e4 });
+  return stdout.trim().length > 0;
+}
 function parseAsr(raw) {
   return responseSchema.parse(raw).result.utterances;
 }
@@ -3389,6 +3396,18 @@ function resolveComparison(doc, episode, cues, comparison) {
 }
 async function alignEpisode(doc, episode, video, scriptPath, config, onAsrComplete) {
   const root = path4.join(path4.dirname(scriptPath), ".review", "asr", `episode-${episode}`);
+  await fs6.mkdir(root, { recursive: true });
+  if (!await hasAudioTrack(video)) {
+    const report2 = {
+      summary: "\u539F\u7247\u6CA1\u6709\u97F3\u8F68\uFF0C\u65E0\u6CD5\u6267\u884C ASR \u6821\u65F6\uFF1B\u7EE7\u7EED\u5BA1\u6838\u753B\u9762\u4E0E\u5267\u672C\u3002",
+      timestamp_corrections: [],
+      missing_dialogue: [],
+      notes: ["ffprobe \u786E\u8BA4\u539F\u7247\u6CA1\u6709\u97F3\u9891\u6D41\u3002\u4FDD\u7559\u73B0\u6709\u53F0\u8BCD\u53CA\u65F6\u95F4\uFF0C\u4E0D\u6839\u636E\u65E0\u97F3\u8F68\u63A8\u65AD\u53F0\u8BCD\u4E0D\u5B58\u5728\uFF1B\u770B\u7247\u6838\u5BF9\u753B\u9762\u4E0E\u5B57\u5E55\uFF0C\u65E0\u6CD5\u4ECE\u753B\u9762\u786E\u8BA4\u7684\u53D1\u58F0\u5185\u5BB9\u660E\u786E\u6807\u4E3A\u8BC1\u636E\u4E0D\u8DB3\u3002"]
+    };
+    const reportDir2 = await fs6.mkdtemp(path4.join(root, "no-audio-"));
+    await fs6.writeFile(path4.join(reportDir2, "report.json"), JSON.stringify(report2, null, 2));
+    return { report: report2, reportDir: reportDir2, usage: [] };
+  }
   const utterances = await episodeAsr(video, root);
   onAsrComplete?.();
   const script = read(doc, String(episode), { timestamps: true });