@lingjingai/scriptctl 0.35.0 → 0.36.1

package/dist/llm/tasks/ingest/video-correct.js

@@ -0,0 +1,144 @@
+// ingest.video-correct task (§10.2 pass4, ported from cezhi-verify 04). Second
+// Gemini pass: video + merged roster (context) + first-pass transcript → flat
+// JSON corrections (name/same_person/speaker/dialogue fixes + notes). Reuses the
+// episode's already-uploaded file. Prompt + schema VERBATIM from the tuned
+// scaffold; Gemini structured output via responseJsonSchema.
+import { CliError, EXIT_RUNTIME } from "../../../common.js";
+export const VIDEO_CORRECT_SYSTEM = "你是短剧剧本的跨集校对器。你会拿到：本集原视频、一份跨集资产花名册(context)、以及第一遍对本集生成的剧本 md。" +
+    "第一遍是单集独立生成的，可能有四类错：①字幕谐音把人名带偏（本名被写成同音近音的别字）；②同一个人跨集用了不同称呼（某集用职务/外貌称呼，花名册里有其规范本名）；③口型与字幕冲突导致对白抄错；" +
+    "④**说话人认错**——第一遍按「画面里此刻是谁的脸」定说话人，但剧里有几个长相相似的角色（尤其一批花白头发、国字脸的中老年男性高管），画面识别会把他们互相认错。" +
+    "对第④类，光看脸救不了（他们本来就像），必须用**剧情逻辑**：" +
+    "台词内容、彼此的称呼（谁叫谁「哥」「老许」「X总」）、场景里的关系（谁在求谁、谁在训谁、谁下跪谁站着、谁是主角谁是对手），" +
+    "这些能锁定『这场戏在场的是哪几个人、每句话是谁说的』——这是跨镜头的剧情推理，比单帧人脸可靠。" +
+    "推理顺序：**先不管第一遍标了谁，只从台词+称呼+剧情关系推断这场有哪些人、谁对谁说话；再回头比对第一遍标的说话人，矛盾的就用 speaker_fixes 纠正。剧情锁定的身份优先于画面外貌。**" +
+    "你的任务：对照原视频画面(人脸/口型/视觉锚点)、剧情逻辑、和 context 花名册，只输出**校正指令**，不要重写整集。没有要改的就返回空数组。";
+const CORRECTION_SCHEMA = {
+    type: "object",
+    additionalProperties: false,
+    required: ["name_fixes", "same_person", "speaker_fixes", "dialogue_fixes", "notes"],
+    properties: {
+        name_fixes: {
+            type: "array",
+            description: "人物命名校正：字幕谐音坑、跨集同人不同称呼",
+            items: {
+                type: "object",
+                additionalProperties: false,
+                required: ["from", "to", "scope", "reason"],
+                properties: {
+                    from: { type: "string", description: "第一遍 md 里用的名字（可能是字幕谐音带偏的，或未听到本名时按职务/外貌起的名）" },
+                    to: { type: "string", description: "校正后的规范名（来自 context 花名册）" },
+                    scope: { type: "string", enum: ["all", "dialogue_only"], description: "all=该名字所有出现都改；dialogue_only=只改对白里的称呼" },
+                    reason: { type: "string", description: "依据：视觉锚点/口型/context 匹配，一句话" },
+                },
+            },
+        },
+        same_person: {
+            type: "array",
+            description: "跨集同人合并判定（本名 vs 职务称呼 vs 听错的名）",
+            items: {
+                type: "object",
+                additionalProperties: false,
+                required: ["names", "reason"],
+                properties: {
+                    names: { type: "array", items: { type: "string" }, description: "判定为同一个人的多个称呼，规范名放第一个" },
+                    reason: { type: "string", description: "依据的视觉锚点，一句话" },
+                },
+            },
+        },
+        speaker_fixes: {
+            type: "array",
+            description: "说话人/在场人校正：第一遍按画面里谁的脸定说话人，但长相相似的角色（一批花白头发老头）会认错。用剧情逻辑（谁对谁说话、称呼、场景关系）纠正标错的人。",
+            items: {
+                type: "object",
+                additionalProperties: false,
+                required: ["ts", "from", "to", "reason"],
+                properties: {
+                    ts: { type: "string", description: "这句台词/动作的时间戳，原样照抄第一遍 md 里那一行的时间戳（MM:SS.sss）" },
+                    from: { type: "string", description: "第一遍标错的说话人/在场人名" },
+                    to: { type: "string", description: "按剧情逻辑应为的正确人名（来自 context 花名册）" },
+                    reason: { type: "string", description: "剧情依据：台词内容/称呼/场景关系为什么锁定这个人，一句话（如：这场是甲向乙下跪求他回来，跪着说'哥对不起你'的只能是甲，第一遍误标成了另一个长相相似的角色）" },
+                },
+            },
+        },
+        dialogue_fixes: {
+            type: "array",
+            description: "对白内容校正：仅限同音谐音错字、串台词、关键数字错。不做称呼规范化——台词里怎么称呼一律保留原话。",
+            items: {
+                type: "object",
+                additionalProperties: false,
+                required: ["ts", "speaker", "wrong", "right", "reason"],
+                properties: {
+                    ts: { type: "string", description: "对白时间戳，原样照抄第一遍 md 里那一行的时间戳（MM:SS.sss）" },
+                    speaker: { type: "string", description: "规范说话人名（此字段可用规范名；但下面的 right 台词正文不许用规范名替换称呼）" },
+                    wrong: { type: "string", description: "第一遍里抄错的对白文本片段" },
+                    right: { type: "string", description: "订正后的对白文本。**只订正同音/谐音错字或串台词**，其余保持原样。严禁把台词里对人物的称呼（某总/某董/某哥/老某）替换成花名册规范名（那个人的本名）——称呼是台词原话，规范名只进 speaker 字段。" },
+                    reason: { type: "string", description: "为什么是抄错：说明是『同音错字订正』（徐→许）还是『串台词/数字错』。若你写的理由是\"规范名/称呼规范化/应称呼其本名\"，说明这是越界，不要提交这条。" },
+                },
+            },
+        },
+        notes: { type: "array", items: { type: "string" }, description: "其他值得记录的问题（如镜头误读、幻觉），不强制修，仅记录" },
+    },
+};
+export function videoCorrectPrompt(episodeNo, contextMd, pass1Md) {
+    return `第 ${episodeNo} 集的校对。
+
+========== 跨集资产 CONTEXT（花名册，规范名以此为准） ==========
+${contextMd}
+
+========== 第一遍对本集生成的剧本 md（待校正） ==========
+${pass1Md}
+
+========== 任务 ==========
+对照上方原视频，逐条检查第一遍 md：
+1. **name_fixes**：本集里有没有被字幕带偏的人名（对口型/看字幕字样判断），或用了「X总」这类称呼而 context 里有对应规范本名的 —— 给出 from→to。
+2. **same_person**：本集出现的人，和 context 花名册里哪个是同一个人（比对视觉锚点：发型/眼镜/脸型/体型/年龄段）。**这是重点，要主动合并，不要因为名字不同就漏判**：
+   - 本集常用职称/敬称（「X总」「X董」「X秘书」）指代的，几乎一定是花名册里已有的某个规范本名 —— 逐个比对锚点，吻合就合并（如 本集某「X总」戴金丝眼镜+花白背头+国字脸 ⇄ 花名册里某本名同锚点 ⇒ 判同人，规范名用那个本名放第一个）。
+   - 判据是**视觉锚点吻合**，不要求名字字面相同。锚点有轻微出入（银丝/金丝眼镜、灰/花白头发）属正常噪声，不影响判定。
+   - 只有当锚点明显冲突（性别/年龄段/发型完全不符）才不合并。宁可主动合并本集的「X总」到花名册规范名，也不要留一堆孤立的「X总」。
+3. **speaker_fixes（重点，治长相相似的老头互认）**：逐场用剧情逻辑核对说话人/在场人：
+   - 先从台词内容+称呼+场景关系推断「这场戏在场的是谁、谁对谁说话」，**不要先信第一遍标的名字**。
+   - 例：某场台词是「我跪下了」「哥对不起你」「我不该听那人的鬼话」——这是一个认错求人的角色（结合上下文是甲向乙下跪求他），这场只有甲+乙，没有别人；如果第一遍把跪着说话的标成了另一个长相相似的第三人，就是认错了，报 from=那个第三人 to=甲。
+   - 判据是**剧情**不是脸：几个长相相近的角色（如一批花白头发、国字脸的中老年男性）画面像，但「谁在这场、谁说这话」由剧情唯一确定。凡第一遍 speaker 与剧情矛盾的，都纠正。
+4. **dialogue_fixes（严格限定，别越界改台词）**：只修两类**明显抄错**：
+   - **同音/谐音错字**：字幕把某个音写成了错字，但读音相同、指的是同一个对象——只把错字订正成对的字（例：某人的姓被写成同音的别字，则把带这个姓的称呼里的错字改回对的姓，如把同音写错的「X总」订正为正确姓氏的「X总」；名字某个字听岔写成近音字，订正回对的字）。订正的是**错字本身**，不是换一种叫法。
+   - **串台词/关键数字错**：整句抄到别处、或金额数字明显抄错。
+   - **铁律：台词内容一个字都不能按花名册去"规范化"。** 人物在台词里**被怎么称呼**（某个「X董」「X哥」「X总」「老X」）是这句台词的原话，即使花名册规范名是他的本名，也**绝对不许**把台词里的称呼替换成规范名——台词里是「X董」就保持「X董」，是「X哥」就保持「X哥」。称呼是谁、规范名是谁，是两回事；规范名只用于 speaker/name 字段，永不进台词正文。
+   - 判据自检：如果你想把 wrong→right 改的是「一种称谓换成另一种称谓」（董→天明、哥→总、总→本名），那就是越界，不要改；只有「同一个音的错字订正」才做。
+5. **notes**：镜头误读、明显幻觉（如把满场的镜头描述成空场）等，记录即可。
+
+只输出校正，不确定的不要瞎改。严格按 schema 返回 JSON。`;
+}
+function correctError(message, received) {
+    return new CliError("INGEST VIDEO CORRECT FAILED", message, {
+        exitCode: EXIT_RUNTIME, errorCode: "INGEST_VIDEO_CORRECT_FAILED", received,
+        nextSteps: ["Inspect the episode error sidecar and rerun `scriptctl ingest`."],
+    });
+}
+const asList = (v) => (Array.isArray(v) ? v : []);
+const str = (v) => (typeof v === "string" ? v : "");
+const strList = (v) => asList(v).map(str).filter(Boolean);
+export async function runVideoCorrect(port, file, episodeNo, contextMd, pass1Md, fps, signal) {
+    const response = await port.generate({
+        file,
+        system: VIDEO_CORRECT_SYSTEM,
+        user: videoCorrectPrompt(episodeNo, contextMd, pass1Md),
+        fps,
+        responseJsonSchema: CORRECTION_SCHEMA,
+        ...(signal ? { signal } : {}),
+    });
+    let raw;
+    try {
+        raw = JSON.parse(response.text);
+    }
+    catch {
+        throw correctError(`episode ${episodeNo}: corrections were not valid JSON`, [response.text.slice(0, 200)]);
+    }
+    return {
+        name_fixes: asList(raw["name_fixes"]).map((x) => { const o = x; return { from: str(o["from"]), to: str(o["to"]), scope: str(o["scope"]) || "all", reason: str(o["reason"]) }; }),
+        same_person: asList(raw["same_person"]).map((x) => { const o = x; return { names: strList(o["names"]), reason: str(o["reason"]) }; }),
+        speaker_fixes: asList(raw["speaker_fixes"]).map((x) => { const o = x; return { ts: str(o["ts"]), from: str(o["from"]), to: str(o["to"]), reason: str(o["reason"]) }; }),
+        dialogue_fixes: asList(raw["dialogue_fixes"]).map((x) => { const o = x; return { ts: str(o["ts"]), speaker: str(o["speaker"]), wrong: str(o["wrong"]), right: str(o["right"]), reason: str(o["reason"]) }; }),
+        notes: strList(raw["notes"]),
+    };
+}
+//# sourceMappingURL=video-correct.js.map

package/dist/llm/tasks/ingest/video-correct.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"video-correct.js","sourceRoot":"","sources":["../../../../src/llm/tasks/ingest/video-correct.ts"],"names":[],"mappings":"AAAA,+EAA+E;AAC/E,8EAA8E;AAC9E,iFAAiF;AACjF,2EAA2E;AAC3E,6DAA6D;AAE7D,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAG5D,MAAM,CAAC,MAAM,oBAAoB,GAC/B,8DAA8D;IAC9D,kGAAkG;IAClG,mFAAmF;IACnF,kCAAkC;IAClC,+DAA+D;IAC/D,iDAAiD;IACjD,oGAAoG;IACpG,6EAA6E,CAAC;AAEhF,MAAM,iBAAiB,GAAe;IACpC,IAAI,EAAE,QAAQ;IACd,oBAAoB,EAAE,KAAK;IAC3B,QAAQ,EAAE,CAAC,YAAY,EAAE,aAAa,EAAE,eAAe,EAAE,gBAAgB,EAAE,OAAO,CAAC;IACnF,UAAU,EAAE;QACV,UAAU,EAAE;YACV,IAAI,EAAE,OAAO;YACb,WAAW,EAAE,uBAAuB;YACpC,KAAK,EAAE;gBACL,IAAI,EAAE,QAAQ;gBACd,oBAAoB,EAAE,KAAK;gBAC3B,QAAQ,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,QAAQ,CAAC;gBAC3C,UAAU,EAAE;oBACV,IAAI,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,2CAA2C,EAAE;oBAClF,EAAE,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,yBAAyB,EAAE;oBAC9D,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,KAAK,EAAE,eAAe,CAAC,EAAE,WAAW,EAAE,sCAAsC,EAAE;oBAC9G,MAAM,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,2BAA2B,EAAE;iBACrE;aACF;SACF;QACD,WAAW,EAAE;YACX,IAAI,EAAE,OAAO;YACb,WAAW,EAAE,8BAA8B;YAC3C,KAAK,EAAE;gBACL,IAAI,EAAE,QAAQ;gBACd,oBAAoB,EAAE,KAAK;gBAC3B,QAAQ,EAAE,CAAC,OAAO,EAAE,QAAQ,CAAC;gBAC7B,UAAU,EAAE;oBACV,KAAK,EAAE,EAAE,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,WAAW,EAAE,sBAAsB,EAAE;oBACxF,MAAM,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,aAAa,EAAE;iBACvD;aACF;SACF;QACD,aAAa,EAAE;YACb,IAAI,EAAE,OAAO;YACb,WAAW,EAAE,4EAA4E;YACzF,KAAK,EAAE;gBACL,IAAI,EAAE,QAAQ;gBACd,oBAAoB,EAAE,KAAK;gBAC3B,QAAQ,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,QAAQ,CAAC;gBACxC,UAAU,EAAE;oBACV,EAAE,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,4CAA4C,EAAE;oBACjF,IAAI,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,gBAAgB,EAAE;oBACvD,EAAE,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,8BAA8B,EAAE;oBACnE,MAAM,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,iFAAiF,EAAE;iBAC3H;aACF;SACF;QACD,cAAc,EAAE;YACd,IAAI,EAAE,OAAO;YACb,WAAW,EAAE,mDAAmD;YAChE,KAAK,EAAE;gBACL,IAAI,EAAE,QAAQ;gBACd,oBAAoB,EAAE,KAAK;gBAC3B,QAAQ,EAAE,CAAC,IAAI,EAAE,SAAS,EAAE,OAAO,EAAE,OAAO,EAAE,QAAQ,CAAC;gBACvD,UAAU,EAAE;oBACV,EAAE,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,sCAAsC,EAAE;oBAC3E,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,4CAA4C,EAAE;oBACtF,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,eAAe,EAAE;oBACvD,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,0GAA0G,EAAE;oBAClJ,MAAM,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,WAAW,EAAE,+EAA+E,EAAE;iBACzH;aACF;SACF;QACD,KAAK,EAAE,EAAE,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,WAAW,EAAE,8BAA8B,EAAE;KACjG;CACF,CAAC;AAEF,MAAM,UAAU,kBAAkB,CAAC,SAAiB,EAAE,SAAiB,EAAE,OAAe;IACtF,OAAO,KAAK,SAAS;;;EAGrB,SAAS;;;EAGT,OAAO;;;;;;;;;;;;;;;;;;;;mCAoB0B,CAAC;AACpC,CAAC;AAUD,SAAS,YAAY,CAAC,OAAe,EAAE,QAAkB;IACvD,OAAO,IAAI,QAAQ,CAAC,6BAA6B,EAAE,OAAO,EAAE;QAC1D,QAAQ,EAAE,YAAY,EAAE,SAAS,EAAE,6BAA6B,EAAE,QAAQ;QAC1E,SAAS,EAAE,CAAC,iEAAiE,CAAC;KAC/E,CAAC,CAAC;AACL,CAAC;AAED,MAAM,MAAM,GAAG,CAAC,CAAU,EAAa,EAAE,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;AACtE,MAAM,GAAG,GAAG,CAAC,CAAU,EAAU,EAAE,CAAC,CAAC,OAAO,CAAC,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;AACrE,MAAM,OAAO,GAAG,CAAC,CAAU,EAAY,EAAE,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC;AAE7E,MAAM,CAAC,KAAK,UAAU,eAAe,CACnC,IAAmB,EACnB,IAAqB,EACrB,SAAiB,EACjB,SAAiB,EACjB,OAAe,EACf,GAAW,EACX,MAAoB;IAEpB,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,QAAQ,CAAC;QACnC,IAAI;QACJ,MAAM,EAAE,oBAAoB;QAC5B,IAAI,EAAE,kBAAkB,CAAC,SAAS,EAAE,SAAS,EAAE,OAAO,CAAC;QACvD,GAAG;QACH,kBAAkB,EAAE,iBAAiB;QACrC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;KAC9B,CAAC,CAAC;IACH,IAAI,GAA4B,CAAC;IACjC,IAAI,CAAC;QACH,GAAG,GAAG,IAAI,CAAC,KAAK,CAAC,QAAQ,CAAC,IAAI,CAA4B,CAAC;IAC7D,CAAC;IAAC,MAAM,CAAC;QACP,MAAM,YAAY,CAAC,WAAW,SAAS,mCAAmC,EAAE,CAAC,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC;IAC7G,CAAC;IACD,OAAO;QACL,UAAU,EAAE,MAAM,CAAC,GAAG,CAAC,YAAY,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,GAAG,MAAM,CAAC,GAAG,CAA4B,CAAC,CAAC,OAAO,EAAE,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,EAAE,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,EAAE,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI,KAAK,EAAE,MAAM,EAAE,GAAG,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;QAC3M,WAAW,EAAE,MAAM,CAAC,GAAG,CAAC,aAAa,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,GAAG,MAAM,CAAC,GAAG,CAA4B,CAAC,CAAC,OAAO,EAAE,KAAK,EAAE,OAAO,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,EAAE,MAAM,EAAE,GAAG,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;QAChK,aAAa,EAAE,MAAM,CAAC,GAAG,CAAC,eAAe,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,GAAG,MAAM,CAAC,GAAG,CAA4B,CAAC,CAAC,OAAO,EAAE,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,EAAE,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,EAAE,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,EAAE,MAAM,EAAE,GAAG,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;QAClM,cAAc,EAAE,MAAM,CAAC,GAAG,CAAC,gBAAgB,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,GAAG,MAAM,CAAC,GAAG,CAA4B,CAAC,CAAC,OAAO,EAAE,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,EAAE,OAAO,EAAE,GAAG,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,EAAE,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,EAAE,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,EAAE,MAAM,EAAE,GAAG,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;QACxO,KAAK,EAAE,OAAO,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC;KAC7B,CAAC;AACJ,CAAC"}

package/dist/llm/tasks/ingest/video-transcribe.d.ts

@@ -0,0 +1,4 @@
+import type { TaskVideoPort, VideoFileHandle } from "../../types.js";
+export declare const VIDEO_TRANSCRIBE_SYSTEM: string;
+export declare function buildTranscribeUser(episodeNo: number): string;
+export declare function runVideoTranscribe(port: TaskVideoPort, file: VideoFileHandle, episodeNo: number, fps: number, signal?: AbortSignal): Promise<string>;

package/dist/llm/tasks/ingest/video-transcribe.js

@@ -0,0 +1,105 @@
+// ingest.video-transcribe task (§10.2 pass1). Gemini turns one episode video
+// into a high-fidelity screenplay markdown: `## 梗概`, then `# 剧本` (scenes with
+// sub-second [MM:SS.sss] timestamps + typed-speaker anchors), then `# 资产` (a
+// character roster carrying cross-episode visual anchors/role — the evidence the
+// merge layer needs). Built on the text MARKDOWN spec, extended with timestamps
+// and per-character anchors. The upload/reuse of the media file is orchestrated
+// by the pipeline (a handle is passed in); this task only generates.
+import { CliError, EXIT_RUNTIME } from "../../../common.js";
+export const VIDEO_TRANSCRIBE_SYSTEM = "你是短剧视频转剧本的分析器。输入是一整集短剧视频（竖屏、1-2 分钟）。" +
+    "你的任务是把这一集**逐镜头、高保真**地转写成一份 markdown 剧本流文档，供转绘 agent 后续出图使用。" +
+    "核心要求：不漏镜头、动作/对白/实体准确、不脑补、时间戳对齐画面。";
+export function buildTranscribeUser(episodeNo) {
+    return `把这一集（第 ${episodeNo} 集）视频转成一份 markdown 剧本流文档。
+
+Output format: 一份 markdown 文档。顺序固定：先 \`## 梗概\`，再 \`# 剧本\`，再 \`# 资产\`。
+不要用 \`\`\` 代码块包裹，不要在文档前后加任何解释性文字。
+
+## 梗概
+<按发生顺序概括本集剧情，200-300 字，纯文本，不用 markdown 结构语法>
+
+# 剧本
+
+## 场景 1
+- 时空: <内|外> <日|夜>
+- 时间: <MM:SS.sss-MM:SS.sss，本场在视频中的起止时间>
+- 地点: <稳定主空间地点名>
+- 角色: <单人角色1>(状态), <角色2>(状态)   ← 状态=整套着装造型，如 (黑色西装)
+- 道具: <需保持一致的道具1>, <道具2>
+
+> [act] [MM:SS.sss] 动作或环境描写（逐镜头，含镜头景别）。**只写画面里发生的动作/环境，绝不把台词字幕抄进来**——短剧每句话底部那条对白字幕是冗余的，属于对应的 [dlg]，act 里一个字都不要出现（详见「画面字幕的取舍」铁律）。
+> [dlg|name] [MM:SS.sss] 角色对白（以口型/音轨为准）。
+> [dlg|name|emotion] [MM:SS.sss] 带情绪标签的对白。
+> [dlg|broadcast:名] [MM:SS.sss] 广播/画外音/主持人等非人物发声。
+> [think|name] [MM:SS.sss] 角色心声/内心独白（画外音但明确是某人所想）。
+
+## 场景 2
+...
+
+# 资产
+
+## 人物
+每个点名或反复出现的具体单人角色，用如下**多行**格式（这是本协议的重点，务必填全）：
+- <角色名>
+  外观: <跨集恒定的身份长相：性别 + 年龄段 + 脸型五官 + 发型发色 + 体型身高 + 固定标记（痣/疤/眼镜/胡须）+ 气质。**不含服饰**，服饰属状态>
+  锚点: <最好认的 1-3 个视觉特征，一句话，用于在别的镜头里重新指认这个人。例：花白背头+金丝眼镜+国字脸>
+  声线: <音色/口音/说话方式，画外或未露脸时的辅助辨识。没有可写「未知」>
+  作用: <这个人在剧情里的身份与关系，例：男主 / 空降副总裁、男主对头 / 男主多年副手>
+
+## 场景
+- <地点名>: <空间类型/规模/风格/主色调/光线/陈设>
+
+## 道具
+- <道具名>: <类别/尺寸/材质/颜色/标志特征>
+
+## 发声源
+- <发声源名>(broadcast|system|prop|location): <声音描述>
+
+## 人物状态
+- <角色名>|<状态名>: <这个状态下的具体可视外观，只写稳定造型，不写动作/情绪/镜头>
+
+规则：
+- 台词以人物**实际开口（口型/音轨）**为准。画面字幕可能是**谐音错字或译文**，与口型冲突时以口型为准，把你判断的正确人名/词写进对白正文即可。
+- **画面字幕的取舍（铁律，最容易犯错，务必照做）**：短剧几乎每句台词底部都压着一条对白字幕，它只是把角色**正在说的这句话**用文字又显示一遍——这类字幕是**对白的冗余复制，必须彻底丢弃**，只保留对应的 [dlg] 行即可，**绝对不要**在任何 act 里写「画面字幕：<某句话>」去复述台词。
+  - 判据（每遇到一条画面文字都过一遍）：这条字幕的内容，是不是差不多就是紧邻的某句台词（某个 [dlg]/[think]）？——**是 ⇒ 直接删，act 里不留任何痕迹**；否 ⇒ 才可能保留。
+  - 正例（该删）：一句台词是「你少管闲事」，画面底部同时显示同样的字幕——只保留那条 [dlg|某人] 你少管闲事，**不要**再写一条 [act] 去注明「画面字幕：你少管闲事」。哪怕字幕和台词有一两个字出入（谐音/漏字），只要显然是同一句，也删。
+  - 反例（该留，写进 act 并注明「画面字幕：xxx」）：**不是任何人此刻在说的话**、承载对白之外新信息的画面文字，才留——人物介绍/身份头衔条、系统/UI 提示文字、时间地点交代（如「三年后」「上海·总部」）、旁白性说明、关键道具上的文字。判别关键：它是「谁正在开口说的台词」还是「画面上一段独立信息」？前者删，后者留。
+  - **人物介绍/姓名头衔条例外**：它承载正式姓名这一新信息，既要写进 act，又要据它把对应角色命名为该正式姓名（见「角色命名」）。
+- 角色命名（重要）：
+  - **命名优先级：书面正式姓名 > 口头喊出的本名 > 纯描述名。** 三者按下面规则取最高可得的一档。
+  - **画面给了书面正式姓名**（人物介绍条/姓名头衔卡/来电显示/工牌/名片/署名等，明确把一个正式姓名标注在某个人身上）→ **就用这个正式姓名当该角色的名**，哪怕全集里所有人只当面喊他职务/敬称（「X总」「X董」「老X」）。书面介绍条是剧方给出的权威身份，优先级高于旁人对他的称呼。判断这条介绍字幕是**压在哪个人身上**（谁的镜头/谁的位置），就把名安到那个人，并在后续该角色的所有 [dlg]/[think] 都用这个正式姓名，保持全集统一。
+  - **听到本名**（对白里明确喊出的姓名）而没有书面姓名条 → 用听到的规范本名（不要用字幕同音近音的误写）。
+  - **既没有书面姓名、也没听到本名** → 只能用**不含姓氏的纯描述名**：按身份或外貌，如「前上司」「下跪的男人」「黑衣女子」「湿发中年男」。并在 锚点 写清体貌，方便第二遍跨集对号。
+  - **严禁凭空捏造带姓氏或像真名的称呼**：没有在视频里真实听到/看到「王总」「李董」「小许」这类称呼时，绝不能自己编一个安上去。捏造的带姓假名会被下游误当成真实人名，污染跨集归并（例：某集把没听到名字的角色随手叫「王总」，其实他是花名册里的别人，假姓一旦落地就很难纠）。宁可用「前上司」这种一眼可见是临时描述的名，也不要编「王总」。
+  - 只有当某个「X总/X董/X姐」称呼是**视频里真实出现的**（有人这么叫他、或字幕这么写），才可以用。
+- \`## 人物\` 只登记具体单人角色；众人/群众/路人用 \`group:名\` 发声或写进 action，不进人物。
+- 状态=整套着装/发型/持续妆造等宏观造型档位，用描述性命名（金色礼服/深灰西装），不用「默认/常态」。情绪、姿态、临时脏污、普通便装不是状态。
+- action 正文逐镜头写实，不要合并镜头、不要脑补画面外信息。
+- 每个 [act]/[dlg]/[think] 都要带 [MM:SS.sss] 时间戳（毫秒三位，精确对齐该镜头/该句起始画面）。
+
+现在开始输出这一集的完整 markdown 文档。`;
+}
+function transcribeError(message, received) {
+    return new CliError("INGEST TRANSCRIBE FAILED", message, {
+        exitCode: EXIT_RUNTIME,
+        errorCode: "INGEST_TRANSCRIBE_FAILED",
+        received,
+        nextSteps: ["Inspect the episode error sidecar and rerun `scriptctl ingest`."],
+    });
+}
+// Transcribe one already-uploaded episode video to markdown. A too-short result
+// is treated as a failed transcription (§12.3: empty responses do not retry).
+export async function runVideoTranscribe(port, file, episodeNo, fps, signal) {
+    const response = await port.generate({
+        file,
+        system: VIDEO_TRANSCRIBE_SYSTEM,
+        user: buildTranscribeUser(episodeNo),
+        fps,
+        ...(signal ? { signal } : {}),
+    });
+    const markdown = response.text.trim();
+    if (markdown.length < 40)
+        throw transcribeError(`episode ${episodeNo}: transcript too short`, [`length: ${markdown.length}`]);
+    return markdown;
+}
+//# sourceMappingURL=video-transcribe.js.map

package/dist/llm/tasks/ingest/video-transcribe.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"video-transcribe.js","sourceRoot":"","sources":["../../../../src/llm/tasks/ingest/video-transcribe.ts"],"names":[],"mappings":"AAAA,6EAA6E;AAC7E,8EAA8E;AAC9E,6EAA6E;AAC7E,iFAAiF;AACjF,gFAAgF;AAChF,gFAAgF;AAChF,qEAAqE;AAErE,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAG5D,MAAM,CAAC,MAAM,uBAAuB,GAClC,sCAAsC;IACtC,6DAA6D;IAC7D,mCAAmC,CAAC;AAEtC,MAAM,UAAU,mBAAmB,CAAC,SAAiB;IACnD,OAAO,UAAU,SAAS;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;0BAmEF,CAAC;AAC3B,CAAC;AAED,SAAS,eAAe,CAAC,OAAe,EAAE,QAAkB;IAC1D,OAAO,IAAI,QAAQ,CAAC,0BAA0B,EAAE,OAAO,EAAE;QACvD,QAAQ,EAAE,YAAY;QACtB,SAAS,EAAE,0BAA0B;QACrC,QAAQ;QACR,SAAS,EAAE,CAAC,iEAAiE,CAAC;KAC/E,CAAC,CAAC;AACL,CAAC;AAED,gFAAgF;AAChF,8EAA8E;AAC9E,MAAM,CAAC,KAAK,UAAU,kBAAkB,CACtC,IAAmB,EACnB,IAAqB,EACrB,SAAiB,EACjB,GAAW,EACX,MAAoB;IAEpB,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,QAAQ,CAAC;QACnC,IAAI;QACJ,MAAM,EAAE,uBAAuB;QAC/B,IAAI,EAAE,mBAAmB,CAAC,SAAS,CAAC;QACpC,GAAG;QACH,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;KAC9B,CAAC,CAAC;IACH,MAAM,QAAQ,GAAG,QAAQ,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC;IACtC,IAAI,QAAQ,CAAC,MAAM,GAAG,EAAE;QAAE,MAAM,eAAe,CAAC,WAAW,SAAS,wBAAwB,EAAE,CAAC,WAAW,QAAQ,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IAC9H,OAAO,QAAQ,CAAC;AAClB,CAAC"}

package/dist/llm/tasks/schemas.d.ts

@@ -0,0 +1,5 @@
+import type { JsonSchema } from "../types.js";
+export declare const EPISODE_TITLE_SCHEMA: JsonSchema;
+export declare const SYNOPSIS_SCHEMA: JsonSchema;
+export declare const SCRIPT_SYNOPSIS_SCHEMA: JsonSchema;
+export declare const STATE_BINDING_SCHEMA: JsonSchema;

package/dist/llm/tasks/schemas.js

@@ -0,0 +1,107 @@
+export const EPISODE_TITLE_SCHEMA = {
+    type: "object",
+    required: ["episode_titles"],
+    additionalProperties: false,
+    properties: {
+        episode_titles: {
+            type: "array",
+            items: {
+                type: "object",
+                required: ["episode", "title"],
+                additionalProperties: false,
+                properties: { episode: { type: "integer" }, title: { type: "string" } },
+            },
+        },
+    },
+};
+export const SYNOPSIS_SCHEMA = {
+    type: "object",
+    required: ["synopsis"],
+    additionalProperties: false,
+    properties: { synopsis: { type: "string" } },
+};
+export const SCRIPT_SYNOPSIS_SCHEMA = {
+    type: "object",
+    required: ["synopsis", "theme", "logline", "style", "main_characters"],
+    additionalProperties: false,
+    properties: {
+        synopsis: { type: "string" },
+        theme: { type: "string" },
+        logline: { type: "string" },
+        style: { type: "string" },
+        main_characters: {
+            type: "array",
+            items: {
+                type: "object",
+                required: ["name", "role"],
+                additionalProperties: false,
+                properties: { name: { type: "string" }, role: { type: "string" } },
+            },
+        },
+    },
+};
+const STATE_REF_SCHEMA = {
+    type: "array",
+    items: {
+        type: "object",
+        required: ["asset_id", "state_id"],
+        additionalProperties: false,
+        properties: { asset_id: { type: "string" }, state_id: { type: ["string", "null"] } },
+    },
+};
+export const STATE_BINDING_SCHEMA = {
+    type: "object",
+    required: ["scenes"],
+    additionalProperties: false,
+    properties: {
+        scenes: {
+            type: "array",
+            items: {
+                type: "object",
+                required: ["scene_id"],
+                additionalProperties: false,
+                properties: {
+                    scene_id: { type: "string" },
+                    actors: {
+                        ...STATE_REF_SCHEMA,
+                        items: {
+                            type: "object", required: ["actor_id", "state_id"], additionalProperties: false,
+                            properties: { actor_id: { type: "string" }, state_id: { type: ["string", "null"] } },
+                        },
+                    },
+                    locations: {
+                        ...STATE_REF_SCHEMA,
+                        items: {
+                            type: "object", required: ["location_id", "state_id"], additionalProperties: false,
+                            properties: { location_id: { type: "string" }, state_id: { type: ["string", "null"] } },
+                        },
+                    },
+                    props: {
+                        ...STATE_REF_SCHEMA,
+                        items: {
+                            type: "object", required: ["prop_id", "state_id"], additionalProperties: false,
+                            properties: { prop_id: { type: "string" }, state_id: { type: ["string", "null"] } },
+                        },
+                    },
+                    state_changes: {
+                        type: "array",
+                        items: {
+                            type: "object",
+                            required: ["action_index", "target_kind", "target_id", "to_state_id"],
+                            additionalProperties: false,
+                            properties: {
+                                action_index: { type: "integer" },
+                                target_kind: { type: "string", enum: ["actor", "location", "prop"] },
+                                target_id: { type: "string" },
+                                from_state_id: { type: ["string", "null"] },
+                                to_state_id: { type: "string" },
+                                effective_from: { type: "string", enum: ["before_action", "after_action"] },
+                            },
+                        },
+                    },
+                },
+            },
+        },
+    },
+};
+//# sourceMappingURL=schemas.js.map

package/dist/llm/tasks/schemas.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"schemas.js","sourceRoot":"","sources":["../../../src/llm/tasks/schemas.ts"],"names":[],"mappings":"AAEA,MAAM,CAAC,MAAM,oBAAoB,GAAe;IAC9C,IAAI,EAAE,QAAQ;IACd,QAAQ,EAAE,CAAC,gBAAgB,CAAC;IAC5B,oBAAoB,EAAE,KAAK;IAC3B,UAAU,EAAE;QACV,cAAc,EAAE;YACd,IAAI,EAAE,OAAO;YACb,KAAK,EAAE;gBACL,IAAI,EAAE,QAAQ;gBACd,QAAQ,EAAE,CAAC,SAAS,EAAE,OAAO,CAAC;gBAC9B,oBAAoB,EAAE,KAAK;gBAC3B,UAAU,EAAE,EAAE,OAAO,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,EAAE,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE;aACxE;SACF;KACF;CACF,CAAC;AAEF,MAAM,CAAC,MAAM,eAAe,GAAe;IACzC,IAAI,EAAE,QAAQ;IACd,QAAQ,EAAE,CAAC,UAAU,CAAC;IACtB,oBAAoB,EAAE,KAAK;IAC3B,UAAU,EAAE,EAAE,QAAQ,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE;CAC7C,CAAC;AAEF,MAAM,CAAC,MAAM,sBAAsB,GAAe;IAChD,IAAI,EAAE,QAAQ;IACd,QAAQ,EAAE,CAAC,UAAU,EAAE,OAAO,EAAE,SAAS,EAAE,OAAO,EAAE,iBAAiB,CAAC;IACtE,oBAAoB,EAAE,KAAK;IAC3B,UAAU,EAAE;QACV,QAAQ,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;QAC5B,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;QACzB,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;QAC3B,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;QACzB,eAAe,EAAE;YACf,IAAI,EAAE,OAAO;YACb,KAAK,EAAE;gBACL,IAAI,EAAE,QAAQ;gBACd,QAAQ,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC;gBAC1B,oBAAoB,EAAE,KAAK;gBAC3B,UAAU,EAAE,EAAE,IAAI,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,IAAI,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE;aACnE;SACF;KACF;CACF,CAAC;AAEF,MAAM,gBAAgB,GAAe;IACnC,IAAI,EAAE,OAAO;IACb,KAAK,EAAE;QACL,IAAI,EAAE,QAAQ;QACd,QAAQ,EAAE,CAAC,UAAU,EAAE,UAAU,CAAC;QAClC,oBAAoB,EAAE,KAAK;QAC3B,UAAU,EAAE,EAAE,QAAQ,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,QAAQ,EAAE,EAAE,IAAI,EAAE,CAAC,QAAQ,EAAE,MAAM,CAAC,EAAE,EAAE;KACrF;CACF,CAAC;AAEF,MAAM,CAAC,MAAM,oBAAoB,GAAe;IAC9C,IAAI,EAAE,QAAQ;IACd,QAAQ,EAAE,CAAC,QAAQ,CAAC;IACpB,oBAAoB,EAAE,KAAK;IAC3B,UAAU,EAAE;QACV,MAAM,EAAE;YACN,IAAI,EAAE,OAAO;YACb,KAAK,EAAE;gBACL,IAAI,EAAE,QAAQ;gBACd,QAAQ,EAAE,CAAC,UAAU,CAAC;gBACtB,oBAAoB,EAAE,KAAK;gBAC3B,UAAU,EAAE;oBACV,QAAQ,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;oBAC5B,MAAM,EAAE;wBACN,GAAG,gBAAgB;wBACnB,KAAK,EAAE;4BACL,IAAI,EAAE,QAAQ,EAAE,QAAQ,EAAE,CAAC,UAAU,EAAE,UAAU,CAAC,EAAE,oBAAoB,EAAE,KAAK;4BAC/E,UAAU,EAAE,EAAE,QAAQ,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,QAAQ,EAAE,EAAE,IAAI,EAAE,CAAC,QAAQ,EAAE,MAAM,CAAC,EAAE,EAAE;yBACrF;qBACF;oBACD,SAAS,EAAE;wBACT,GAAG,gBAAgB;wBACnB,KAAK,EAAE;4BACL,IAAI,EAAE,QAAQ,EAAE,QAAQ,EAAE,CAAC,aAAa,EAAE,UAAU,CAAC,EAAE,oBAAoB,EAAE,KAAK;4BAClF,UAAU,EAAE,EAAE,WAAW,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,QAAQ,EAAE,EAAE,IAAI,EAAE,CAAC,QAAQ,EAAE,MAAM,CAAC,EAAE,EAAE;yBACxF;qBACF;oBACD,KAAK,EAAE;wBACL,GAAG,gBAAgB;wBACnB,KAAK,EAAE;4BACL,IAAI,EAAE,QAAQ,EAAE,QAAQ,EAAE,CAAC,SAAS,EAAE,UAAU,CAAC,EAAE,oBAAoB,EAAE,KAAK;4BAC9E,UAAU,EAAE,EAAE,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,QAAQ,EAAE,EAAE,IAAI,EAAE,CAAC,QAAQ,EAAE,MAAM,CAAC,EAAE,EAAE;yBACpF;qBACF;oBACD,aAAa,EAAE;wBACb,IAAI,EAAE,OAAO;wBACb,KAAK,EAAE;4BACL,IAAI,EAAE,QAAQ;4BACd,QAAQ,EAAE,CAAC,cAAc,EAAE,aAAa,EAAE,WAAW,EAAE,aAAa,CAAC;4BACrE,oBAAoB,EAAE,KAAK;4BAC3B,UAAU,EAAE;gCACV,YAAY,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE;gCACjC,WAAW,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,OAAO,EAAE,UAAU,EAAE,MAAM,CAAC,EAAE;gCACpE,SAAS,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;gCAC7B,aAAa,EAAE,EAAE,IAAI,EAAE,CAAC,QAAQ,EAAE,MAAM,CAAC,EAAE;gCAC3C,WAAW,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE;gCAC/B,cAAc,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,eAAe,EAAE,cAAc,CAAC,EAAE;6BAC5E;yBACF;qBACF;iBACF;aACF;SACF;KACF;CACF,CAAC"}

package/dist/llm/types.d.ts

@@ -0,0 +1,97 @@
+export type LlmProtocol = "anthropic" | "openai-compatible" | "google-gemini";
+export type LlmTaskId = "direct.episode-titles" | "direct.batch-extract" | "direct.episode-synopsis" | "direct.synopsis-reduction" | "direct.script-synopsis" | "direct.asset-grouping" | "direct.asset-curation" | "direct.asset-curation-audit" | "direct.location-canonicalization" | "direct.state-curation" | "direct.state-binding" | "metadata.global" | "metadata.descriptions" | "synopsis.chunk" | "synopsis.episode" | "synopsis.arc" | "synopsis.script-overview" | "ingest.text-normalize" | "ingest.mention-resolve" | "ingest.asset-extract" | "ingest.asset-curate" | "ingest.asset-recall" | "ingest.asset-refine" | "ingest.state-reconcile" | "ingest.speaker-audit" | "ingest.video-transcribe" | "ingest.video-correct" | "ingest.roster-recall" | "ingest.roster-refine" | "ingest.roster-canon" | "ingest.roster-remerge" | "ingest.state-merge";
+export type LlmTaskMode = "text" | "tool" | "video";
+export type JsonSchema = Record<string, unknown>;
+export interface TextGenerationRequest {
+    system: string;
+    user: string;
+    maxOutputTokens: number;
+    signal?: AbortSignal;
+}
+export interface ToolDefinition {
+    name: string;
+    description: string;
+    inputSchema: JsonSchema;
+}
+export interface ToolGenerationRequest extends TextGenerationRequest {
+    tool: ToolDefinition;
+}
+export interface TextGenerationResponse {
+    text: string;
+    finishReason: string;
+    attempts: number;
+    raw?: unknown;
+}
+export interface ToolGenerationResponse {
+    input: unknown;
+    finishReason: string;
+    attempts: number;
+    raw?: unknown;
+}
+export interface LlmAdapter {
+    generateText(request: TextGenerationRequest): Promise<TextGenerationResponse>;
+    generateTool(request: ToolGenerationRequest): Promise<ToolGenerationResponse>;
+}
+export interface VideoFileHandle {
+    name: string;
+    uri: string;
+    mimeType: string;
+}
+export interface VideoGenerationRequest {
+    file: VideoFileHandle;
+    system?: string;
+    user: string;
+    fps: number;
+    maxOutputTokens: number;
+    responseJsonSchema?: JsonSchema;
+    signal?: AbortSignal;
+}
+export interface VideoGenerationResponse {
+    text: string;
+    finishReason: string;
+    attempts: number;
+    raw?: unknown;
+}
+export interface VideoUnderstandingPort {
+    upload(localPath: string, mimeType: string, signal?: AbortSignal): Promise<VideoFileHandle>;
+    generate(request: VideoGenerationRequest): Promise<VideoGenerationResponse>;
+    delete(handle: VideoFileHandle, signal?: AbortSignal): Promise<void>;
+}
+export interface TaskVideoPort {
+    upload(localPath: string, mimeType: string, signal?: AbortSignal): Promise<VideoFileHandle>;
+    generate(request: Omit<VideoGenerationRequest, "maxOutputTokens">): Promise<VideoGenerationResponse>;
+    delete(handle: VideoFileHandle, signal?: AbortSignal): Promise<void>;
+}
+export interface LlmRouteInfo {
+    task: LlmTaskId;
+    profile: string;
+    provider: string;
+    protocol: LlmProtocol;
+    model: string;
+    fingerprint: string;
+}
+export interface LlmRouteUsage extends LlmRouteInfo {
+    calls: number;
+    reused: number;
+    attempts: number;
+}
+export interface LlmTaskExecutor {
+    generateText(task: LlmTaskId, request: Omit<TextGenerationRequest, "maxOutputTokens">): Promise<TextGenerationResponse>;
+    generateTool(task: LlmTaskId, request: Omit<ToolGenerationRequest, "maxOutputTokens">): Promise<ToolGenerationResponse>;
+    videoPort(task: LlmTaskId): TaskVideoPort;
+    route(task: LlmTaskId): LlmRouteInfo;
+    routes(): LlmRouteInfo[];
+    recordReuse(task: LlmTaskId): void;
+    usage(): LlmRouteUsage[];
+    readonly configSource: "builtin" | "file";
+}
+export type LlmErrorCode = "auth" | "endpoint" | "timeout" | "rate_limited" | "request_failed" | "truncated" | "content_filtered" | "empty_response" | "invalid_response";
+export declare class LlmTransportError extends Error {
+    readonly code: LlmErrorCode;
+    readonly retryable: boolean;
+    readonly details: string[];
+    constructor(code: LlmErrorCode, message: string, options?: {
+        retryable?: boolean;
+        details?: string[];
+    });
+}

package/dist/llm/types.js

@@ -0,0 +1,13 @@
+export class LlmTransportError extends Error {
+    code;
+    retryable;
+    details;
+    constructor(code, message, options = {}) {
+        super(message);
+        this.name = "LlmTransportError";
+        this.code = code;
+        this.retryable = options.retryable ?? false;
+        this.details = options.details ?? [];
+    }
+}
+//# sourceMappingURL=types.js.map

package/dist/llm/types.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.js","sourceRoot":"","sources":["../../src/llm/types.ts"],"names":[],"mappings":"AA0JA,MAAM,OAAO,iBAAkB,SAAQ,KAAK;IACjC,IAAI,CAAe;IACnB,SAAS,CAAU;IACnB,OAAO,CAAW;IAE3B,YAAY,IAAkB,EAAE,OAAe,EAAE,UAAuD,EAAE;QACxG,KAAK,CAAC,OAAO,CAAC,CAAC;QACf,IAAI,CAAC,IAAI,GAAG,mBAAmB,CAAC;QAChC,IAAI,CAAC,IAAI,GAAG,IAAI,CAAC;QACjB,IAAI,CAAC,SAAS,GAAG,OAAO,CAAC,SAAS,IAAI,KAAK,CAAC;QAC5C,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC,OAAO,IAAI,EAAE,CAAC;IACvC,CAAC;CACF"}

package/dist/usecases/doctor.d.ts

@@ -1,2 +1,2 @@
 import { type Report } from "../common.js";
-export declare function commandDoctor(_opts?: Record<string, unknown>): [Report, number];
+export declare function commandDoctor(opts?: Record<string, unknown>): [Report, number];