@adminide-stack/yantra-mobile 12.0.129-alpha.0 → 12.0.129-alpha.1

package/lib/features/canvas/videoEditor/agent/cutTools.js

@@ -1,4 +1,4 @@
-import {removeClip,moveClip,updateClip,clipFromColor,textClipAt}from'../editorModel.js';import {sceneSpans}from'../sceneLayers.js';var __defProp = Object.defineProperty;
+import {removeClip,moveClip,updateClip,clipFromColor,textClipAt,DEFAULT_IMAGE_SEC,clipFromItem}from'../editorModel.js';import {sceneSpans,sceneAtSec}from'../sceneLayers.js';import {captionsForClip,placeCaptions,transcribeMedia}from'../captions.js';import {STOCK_VOICES,speakLine,composeSoundEffect,composeMusic}from'../audioServices.js';import {probeAudioSec}from'../revoice.js';var __defProp = Object.defineProperty;
 var __defProps = Object.defineProperties;
 var __getOwnPropDescs = Object.getOwnPropertyDescriptors;
 var __getOwnPropSymbols = Object.getOwnPropertySymbols;
@@ -158,7 +158,149 @@ const CUT_AGENT_TOOLS = [{
       }
     }
   }
-}];
+}, ...MEDIA_TOOL_SCHEMAS()];
+function tool(name, description, properties, required = []) {
+  return {
+    type: "function",
+    function: {
+      name,
+      description,
+      parameters: {
+        type: "object",
+        properties,
+        required
+      }
+    }
+  };
+}
+function sceneEntry() {
+  return {
+    type: "object",
+    properties: {
+      asset_id: {
+        type: "string",
+        description: "Library item or timeline clip id (video or image) to cut from. Omit for a scene to film."
+      },
+      in_sec: {
+        type: "number",
+        description: "Start inside the source clip, seconds. Omit for the head."
+      },
+      out_sec: {
+        type: "number",
+        description: "End inside the source clip, seconds. For a still: how long it holds."
+      },
+      label: {
+        type: "string",
+        description: "Short scene name shown on the timeline."
+      },
+      shot: {
+        type: "string",
+        description: "A scene to FILM instead of cutting from the library: subject, setting, action, light, camera move. Picture only. Then call film_scenes."
+      },
+      seconds: {
+        type: "number",
+        description: "How long a scene to film plays (default 6, 2-15)."
+      }
+    }
+  };
+}
+const MEDIA_TOOL_NAMES = ["analyze_footage", "read_document", "get_transcript", "search_dialogue", "import_link", "set_cut", "add_music", "add_sound_effect", "add_voiceover", "add_captions", "set_format"];
+function MEDIA_TOOL_SCHEMAS() {
+  return [tool("analyze_footage", "Learn the dialogue (timed transcript) and, when available, what each library item or clip shows. Cached per item: analyzing again is free. Call before cutting footage you have not seen.", {
+    asset_ids: {
+      type: "array",
+      items: {
+        type: "string"
+      },
+      description: "Only these items; omit for every item not yet analyzed."
+    },
+    visual: {
+      type: "boolean",
+      description: "Describe what the picture shows (default true)."
+    }
+  }), tool("read_document", "The full text of an uploaded document (screenplay, shot list, brief).", {
+    document_id: {
+      type: "string"
+    }
+  }, ["document_id"]), tool("get_transcript", "Every dialogue cue of one item with source timings, once it has been analyzed.", {
+    asset_id: {
+      type: "string"
+    }
+  }, ["asset_id"]), tool("search_dialogue", "Find where words are spoken across the analyzed media: returns the best-matching cues with their item id and source timings.", {
+    query: {
+      type: "string",
+      description: "Words or a line of dialogue to look for."
+    },
+    limit: {
+      type: "number",
+      description: "Max results (default 12)."
+    }
+  }, ["query"]), tool("import_link", "Take a web link into the library: a direct picture, video or sound file, or a page (its preview picture or video). Links in the user's current message are already taken in; use this for a link from an earlier message.", {
+    url: {
+      type: "string"
+    }
+  }, ["url"]), tool("set_cut", 'Lay scenes on the timeline back to back. Each scene is cut from library footage (asset_id) or is a shot to film (shot + seconds). mode "replace" swaps the whole cut; "append" adds after the last scene. Returns the new scene ids.', {
+    scenes: {
+      type: "array",
+      items: sceneEntry(),
+      description: "Scenes in play order."
+    },
+    mode: {
+      type: "string",
+      enum: ["replace", "append"],
+      description: 'Default "replace".'
+    }
+  }, ["scenes"]), tool("add_music", "Generate a music bed from a mood description and lay it under the cut (mixed under the other audio). Takes a minute or two.", {
+    prompt: {
+      type: "string",
+      description: "Mood, genre, instrumentation, tempo, build."
+    },
+    duration_sec: {
+      type: "number",
+      description: "Length wanted (default: the cut's length, max 180)."
+    },
+    at_sec: {
+      type: "number",
+      description: "Timeline second it starts (default 0)."
+    }
+  }, ["prompt"]), tool("add_sound_effect", "Generate one sound effect (a hit, riser, whoosh, ambience) and place it at a timeline second.", {
+    prompt: {
+      type: "string"
+    },
+    at_sec: {
+      type: "number"
+    },
+    duration_sec: {
+      type: "number",
+      description: "Default 2, max 20."
+    }
+  }, ["prompt", "at_sec"]), tool("add_voiceover", "Speak a line with a synthetic voice and place it at a timeline second; it moves with the scene it lands in.", {
+    text: {
+      type: "string"
+    },
+    at_sec: {
+      type: "number",
+      description: "Default: after the last voiceover line, else 0."
+    },
+    voice: {
+      type: "string",
+      enum: STOCK_VOICES.map((v) => v.id),
+      description: "Default: the studio's first voice."
+    }
+  }, ["text"]), tool("add_captions", "Captions from the dialogue of the given scenes (default: every video scene); transcribes what is not analyzed yet.", {
+    scene_ids: {
+      type: "array",
+      items: {
+        type: "string"
+      }
+    }
+  }), tool("set_format", "Output aspect: 16:9, 9:16, 1:1, 4:5, or auto (source).", {
+    aspect: {
+      type: "string",
+      enum: ["16:9", "9:16", "1:1", "4:5", "auto"]
+    }
+  }, ["aspect"])];
+}
 async function executeCutTool(name, rawArgs, ctx) {
   var _a, _b;
   const args = rawArgs && typeof rawArgs === "object" ? rawArgs : {};
@@ -321,9 +463,571 @@ async function executeCutTool(name, rawArgs, ctx) {
       content: `filming ${ids.length} scene(s); each lands in place in a few minutes`
     };
   }
+  if (MEDIA_TOOL_NAMES.includes(name)) {
+    try {
+      return await executeMediaTool(name, args, ctx);
+    } catch (err) {
+      const message = err instanceof Error ? err.message : String(err);
+      const kind = err instanceof ToolInputError ? "bad arguments" : "failed";
+      return {
+        ok: false,
+        label: `${name.replace(/_/g, " ")} ${kind}`,
+        detail: message.slice(0, 80),
+        content: `error (${kind}): ${message}`
+      };
+    }
+  }
   return {
     ok: false,
     label: name,
     content: `Unknown tool ${name}`
   };
-}export{CUT_AGENT_TOOLS,SHOT_SEC_RANGE,executeCutTool};//# sourceMappingURL=cutTools.js.map
+}
+class ToolInputError extends Error {
+}
+function str(args, key, required = false) {
+  const v = args[key];
+  if (v == null || v === "") {
+    if (required) throw new ToolInputError(`missing "${key}"`);
+    return void 0;
+  }
+  if (typeof v !== "string") throw new ToolInputError(`"${key}" must be a string`);
+  return v;
+}
+function num(args, key) {
+  const v = args[key];
+  if (v == null || v === "") return void 0;
+  const n = typeof v === "number" ? v : Number(v);
+  if (!Number.isFinite(n)) throw new ToolInputError(`"${key}" must be a number of seconds`);
+  return n;
+}
+function strList(args, key, required = false) {
+  const v = args[key];
+  if (v == null) {
+    if (required) throw new ToolInputError(`missing "${key}"`);
+    return void 0;
+  }
+  if (!Array.isArray(v)) throw new ToolInputError(`"${key}" must be an array of ids`);
+  return v.map((x) => String(x));
+}
+const secs = (n) => `${n.toFixed(1)}s`;
+const plural = (n, word) => `${n} ${word}${n === 1 ? "" : "s"}`;
+const excerpt = (s, n) => s.length > n ? `${s.slice(0, n - 1)}\u2026` : s;
+function withNotes(text, notes) {
+  return notes.length ? `${text}
+Notes: ${notes.join("; ")}` : text;
+}
+const DOCUMENT_CHAR_CAP = 6e4;
+const TRANSCRIPT_CUE_CAP = 400;
+const DEFAULT_MUSIC_SEC = 60;
+const MAX_MUSIC_SEC = 180;
+const DEFAULT_SFX_SEC = 2;
+const FORMAT_ASPECTS = {
+  "16:9": "16:9",
+  "9:16": "9:16",
+  "1:1": "1:1",
+  "4:5": "4:5",
+  auto: null
+};
+function agentLibrary(ctx) {
+  var _a, _b;
+  const out = [...(_b = (_a = ctx.library) == null ? void 0 : _a.call(ctx)) != null ? _b : []];
+  const seen = new Set(out.map((i) => i.id));
+  for (const c of ctx.getTimeline().clips) {
+    if (c.fillColor || seen.has(c.id)) continue;
+    out.push({
+      id: c.id,
+      type: c.type,
+      src: c.src,
+      name: c.name
+    });
+  }
+  return out;
+}
+function needRun(ctx, what) {
+  if (!ctx.runAction) throw new Error(`${what} is not available here`);
+  return ctx.runAction;
+}
+function cache(ctx) {
+  if (!ctx.analyses) ctx.analyses = /* @__PURE__ */ new Map();
+  return ctx.analyses;
+}
+function cutSec(t, sourceSec) {
+  const spans = sceneSpans(t.clips, sourceSec);
+  return spans.length ? spans[spans.length - 1].endSec : 0;
+}
+let audioCounter = 0;
+function sceneAudioAt(t, audio, sourceSec) {
+  audioCounter += 1;
+  const scene = sceneAtSec(sceneSpans(t.clips, sourceSec), audio.startSec);
+  return __spreadValues(__spreadProps(__spreadValues({
+    id: `agent-audio-${Date.now().toString(36)}-${audioCounter}`
+  }, audio), {
+    startSec: Math.round(audio.startSec * 1e3) / 1e3
+  }), scene ? {
+    sceneId: scene.id
+  } : {});
+}
+async function analyzeItem(item, ctx, visual) {
+  var _a, _b, _c, _d, _e;
+  const seed = cache(ctx).get(item.src);
+  const problems = [];
+  let transcript = (_a = seed == null ? void 0 : seed.transcript) != null ? _a : [];
+  const timed = item.type === "video" || item.type === "audio";
+  if (timed && !seed) {
+    if (ctx.runAction) {
+      try {
+        transcript = await transcribeMedia(ctx.runAction, item.src);
+      } catch (err) {
+        problems.push(`transcription failed: ${err instanceof Error ? err.message : String(err)}`);
+      }
+    } else problems.push("transcription is not available here");
+  }
+  const clip = ctx.getTimeline().clips.find((c) => c.id === item.id);
+  const durationSec = (seed == null ? void 0 : seed.durationSec) || (clip ? clip.type === "image" ? (_b = clip.durationSec) != null ? _b : DEFAULT_IMAGE_SEC : (_e = (_d = (_c = ctx.sourceSec) == null ? void 0 : _c.call(ctx)[clip.id]) != null ? _d : clip.trimEndSec) != null ? _e : 0 : 0) || (transcript.length ? transcript[transcript.length - 1].end : 0);
+  let description = seed == null ? void 0 : seed.description;
+  if (visual && ctx.describe && !description && item.type !== "audio") {
+    try {
+      let picture = item.type === "image" ? item.src : null;
+      if (!picture && ctx.runAction) {
+        const r = await ctx.runAction("yantra-video-editing", "YANTRA_VIDEO_EDIT_CONTACT_SHEET", {
+          input_url: item.src,
+          cols: 3,
+          rows: 2,
+          thumb_width: 320
+        });
+        const root = r.data && typeof r.data === "object" ? r.data : r;
+        picture = typeof root.output_url === "string" && root.output_url ? root.output_url : null;
+      }
+      if (picture) {
+        description = await ctx.describe(picture, item.type === "image" ? "This is a still image from a video project." : "This is a contact sheet of six frames from one video clip, left to right then top to bottom, in time order.");
+      } else problems.push("no frames could be rendered for the visual pass");
+    } catch (err) {
+      problems.push(`visual pass failed: ${err instanceof Error ? err.message : String(err)}`);
+    }
+  }
+  const analysis = __spreadValues({
+    durationSec,
+    transcript
+  }, description ? {
+    description
+  } : {});
+  cache(ctx).set(item.src, analysis);
+  return {
+    analysis,
+    problems
+  };
+}
+function summarize(item, an, problems) {
+  const bits = [`id=${item.id} "${item.name}" (${item.type})`];
+  if (an.durationSec > 0) bits.push(secs(an.durationSec));
+  if (an.transcript.length) bits.push(`dialogue: "${excerpt(an.transcript.map((c) => c.text).join(" "), 220)}"`);
+  else if (item.type !== "image") bits.push("no speech");
+  if (an.description) bits.push(`looks: ${an.description}`);
+  if (problems.length) bits.push(`(${problems.join("; ")})`);
+  return `- ${bits.join(" \xB7 ")}`;
+}
+const tokensOf = (s) => {
+  var _a;
+  return (_a = s.toLowerCase().match(/[a-z0-9']{3,}/g)) != null ? _a : [];
+};
+function clipForEntry(entry, i, library, timeline) {
+  var _a, _b, _c, _d;
+  const assetId = (_a = str(entry, "asset_id")) != null ? _a : str(entry, "id");
+  const shot = (_b = str(entry, "shot")) == null ? void 0 : _b.trim();
+  const label = str(entry, "label");
+  if (!assetId) {
+    if (!shot) throw new ToolInputError(`scenes[${i}] needs an asset_id from the library or a shot to film`);
+    const asked = (_c = num(entry, "seconds")) != null ? _c : SHOT_SEC_RANGE.default;
+    const seconds = Math.min(SHOT_SEC_RANGE.max, Math.max(SHOT_SEC_RANGE.min, asked));
+    return {
+      clip: __spreadProps(__spreadValues({}, clipFromColor("#101014")), {
+        name: (label != null ? label : shot).slice(0, 32),
+        shot,
+        durationSec: seconds
+      })
+    };
+  }
+  const onTimeline = timeline.clips.find((c) => c.id === assetId);
+  const item = library.find((x) => x.id === assetId);
+  if (!onTimeline && !item) return {
+    error: `scenes[${i}]: no library item "${assetId}"`
+  };
+  if ((item == null ? void 0 : item.type) === "audio") return {
+    error: `scenes[${i}]: "${assetId}" is audio, not a picture`
+  };
+  const base = onTimeline ? __spreadValues(__spreadValues(__spreadValues({}, clipFromItem({
+    id: onTimeline.itemId,
+    type: onTimeline.type,
+    src: onTimeline.src,
+    generatedAssetName: onTimeline.name
+  })), onTimeline.fillColor ? {
+    fillColor: onTimeline.fillColor
+  } : {}), onTimeline.hasAudio === false ? {
+    hasAudio: false
+  } : {}) : clipFromItem({
+    id: item.id,
+    type: item.type,
+    src: item.src,
+    generatedAssetName: item.name
+  });
+  const inSec = Math.max(0, (_d = num(entry, "in_sec")) != null ? _d : 0);
+  const outSec = num(entry, "out_sec");
+  const named = label ? {
+    name: label.slice(0, 32)
+  } : {};
+  if (base.type === "image") {
+    const hold = outSec !== void 0 ? Math.max(0.5, outSec - inSec) : DEFAULT_IMAGE_SEC;
+    return {
+      clip: __spreadProps(__spreadValues(__spreadValues({}, base), named), {
+        durationSec: hold
+      })
+    };
+  }
+  if (outSec !== void 0 && outSec <= inSec) return {
+    error: `scenes[${i}]: out_sec must be after in_sec`
+  };
+  return {
+    clip: __spreadProps(__spreadValues(__spreadValues({}, base), named), {
+      trimStartSec: inSec,
+      trimEndSec: outSec != null ? outSec : null
+    })
+  };
+}
+async function executeMediaTool(name, args, ctx) {
+  var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w;
+  const sourceSec = (_a = ctx.sourceSec) == null ? void 0 : _a.call(ctx);
+  switch (name) {
+    case "analyze_footage": {
+      const library = agentLibrary(ctx);
+      const ids = strList(args, "asset_ids");
+      const visual = typeof args.visual === "boolean" ? args.visual : true;
+      const missing = [];
+      const targets = ids ? ids.flatMap((id) => {
+        const a = library.find((x) => x.id === id);
+        if (!a) missing.push(`no item "${id}"`);
+        return a ? [a] : [];
+      }) : library.filter((a) => {
+        const an = cache(ctx).get(a.src);
+        return !an || visual && ctx.describe && !an.description && a.type !== "audio";
+      });
+      if (!targets.length) {
+        return {
+          ok: true,
+          label: "Library already analyzed",
+          content: withNotes("Everything in the library is already analyzed.", missing)
+        };
+      }
+      const lines = [];
+      let spoken = 0;
+      for (let i = 0; i < targets.length; i += 1) {
+        (_b = ctx.progress) == null ? void 0 : _b.call(ctx, `Scanning ${i + 1}/${targets.length}\u2026`);
+        const {
+          analysis,
+          problems
+        } = await analyzeItem(targets[i], ctx, visual);
+        if (analysis.transcript.length) spoken += 1;
+        lines.push(summarize(targets[i], analysis, problems));
+      }
+      return {
+        ok: true,
+        label: `Scanned ${plural(targets.length, "clip")}`,
+        detail: `${spoken} with dialogue`,
+        content: withNotes(`Analyzed ${plural(targets.length, "item")}:
+${lines.join("\n")}`, missing)
+      };
+    }
+    case "read_document": {
+      const id = str(args, "document_id", true);
+      const docs = (_d = (_c = ctx.documents) == null ? void 0 : _c.call(ctx)) != null ? _d : [];
+      const doc = (_e = docs.find((d) => d.id === id)) != null ? _e : docs.find((d) => d.name.toLowerCase() === id.toLowerCase());
+      if (!doc) throw new ToolInputError(`no document "${id}"; available: ${docs.map((d) => `${d.id} (${d.name})`).join(", ") || "none"}`);
+      const cut = doc.text.length > DOCUMENT_CHAR_CAP;
+      return {
+        ok: true,
+        label: `Read ${doc.name}`,
+        content: `${doc.name}${cut ? ` (first ${DOCUMENT_CHAR_CAP} of ${doc.text.length} chars)` : ""}:
+${doc.text.slice(0, DOCUMENT_CHAR_CAP)}`
+      };
+    }
+    case "get_transcript": {
+      const id = str(args, "asset_id", true);
+      const item = agentLibrary(ctx).find((x) => x.id === id);
+      if (!item) throw new ToolInputError(`no library item "${id}"`);
+      const cues = (_f = cache(ctx).get(item.src)) == null ? void 0 : _f.transcript;
+      if (!cues) return {
+        ok: true,
+        label: `No transcript for ${item.name}`,
+        content: `"${item.name}" is not analyzed yet - call analyze_footage.`
+      };
+      if (!cues.length) return {
+        ok: true,
+        label: `No speech in ${item.name}`,
+        content: `"${item.name}" has no speech.`
+      };
+      const lines = cues.slice(0, TRANSCRIPT_CUE_CAP).map((c) => `[${c.start.toFixed(1)}-${c.end.toFixed(1)}] ${c.text}`);
+      return {
+        ok: true,
+        label: `Read transcript of ${item.name}`,
+        content: `${item.name} (${plural(cues.length, "cue")}):
+${lines.join("\n")}`
+      };
+    }
+    case "search_dialogue": {
+      const query = str(args, "query", true);
+      const limit = Math.max(1, Math.min(40, (_g = num(args, "limit")) != null ? _g : 12));
+      const words = new Set(tokensOf(query));
+      const needle = query.toLowerCase().trim();
+      const hits = [];
+      let searched = 0;
+      for (const item of agentLibrary(ctx)) {
+        const cues = (_h = cache(ctx).get(item.src)) == null ? void 0 : _h.transcript;
+        if (!(cues == null ? void 0 : cues.length)) continue;
+        searched += 1;
+        cues.forEach((c, i) => {
+          var _a2, _b2, _c2, _d2;
+          const text = c.text.toLowerCase();
+          let score = 0;
+          for (const w of tokensOf(text)) if (words.has(w)) score += 1;
+          if (needle && text.includes(needle)) score += 5;
+          if (!score) return;
+          const before = (_b2 = (_a2 = cues[i - 1]) == null ? void 0 : _a2.text) != null ? _b2 : "";
+          const after = (_d2 = (_c2 = cues[i + 1]) == null ? void 0 : _c2.text) != null ? _d2 : "";
+          hits.push({
+            score: score / Math.max(1, words.size),
+            line: `- ${item.id} "${item.name}" [${c.start.toFixed(1)}-${c.end.toFixed(1)}] "${c.text}"${before || after ? ` (context: \u2026${excerpt(before, 60)} | ${excerpt(after, 60)}\u2026)` : ""}`
+          });
+        });
+      }
+      if (!searched) return {
+        ok: true,
+        label: "Nothing to search",
+        content: "No analyzed dialogue yet - call analyze_footage first."
+      };
+      hits.sort((a, b) => b.score - a.score);
+      const top = hits.slice(0, limit);
+      return {
+        ok: true,
+        label: `Searched dialogue for "${excerpt(query, 40)}"`,
+        detail: plural(top.length, "match"),
+        content: top.length ? `${plural(top.length, "match")} for "${query}":
+${top.map((h) => h.line).join("\n")}` : `Nothing in the dialogue matches "${query}".`
+      };
+    }
+    case "import_link": {
+      const url = str(args, "url", true).trim();
+      const known = agentLibrary(ctx).find((a) => a.sourceUrl === url);
+      if (known) return {
+        ok: true,
+        label: `Already have ${known.name}`,
+        content: `Already in the library: ${known.type} id=${known.id} "${known.name}".`
+      };
+      if (!ctx.importLink) throw new Error("taking in links is not available here");
+      (_i = ctx.progress) == null ? void 0 : _i.call(ctx, "Opening the link\u2026");
+      const got = await ctx.importLink(url);
+      let looks = "";
+      if (got.type === "image" && ctx.describe) {
+        const {
+          analysis
+        } = await analyzeItem(got, ctx, true);
+        if (analysis.description) looks = ` \xB7 looks: ${analysis.description}`;
+      }
+      return {
+        ok: true,
+        label: `Took in ${got.name}`,
+        content: `Took in the link as ${got.type} id=${got.id} "${got.name}"${looks}.`
+      };
+    }
+    case "set_cut": {
+      const raw = args.scenes;
+      if (!Array.isArray(raw) || !raw.length) throw new ToolInputError('"scenes" must be a non-empty array');
+      const mode = str(args, "mode") === "append" ? "append" : "replace";
+      const t = ctx.getTimeline();
+      const library = agentLibrary(ctx);
+      const notes = [];
+      const made = [];
+      raw.forEach((e, i) => {
+        const entry = e && typeof e === "object" && !Array.isArray(e) ? e : {};
+        const r = clipForEntry(entry, i, library, t);
+        if ("error" in r) notes.push(`skipped ${r.error}`);
+        else made.push(r.clip);
+      });
+      if (!made.length) return {
+        ok: false,
+        label: "Nothing placed",
+        content: withNotes("Nothing was placed.", notes)
+      };
+      const next = __spreadProps(__spreadValues({}, t), {
+        clips: mode === "replace" ? made : [...t.clips, ...made]
+      });
+      ctx.setTimeline(next);
+      const toFilm = made.filter((c) => c.fillColor && c.shot).length;
+      return {
+        ok: true,
+        label: `${mode === "append" ? "Appended" : "Placed"} ${plural(made.length, "scene")}`,
+        detail: `cut is ${secs(cutSec(next, sourceSec))}`,
+        content: withNotes(`Placed ${plural(made.length, "scene")} (${mode}): ${JSON.stringify(made.map((c) => c.id))}.${toFilm ? ` ${plural(toFilm, "scene")} wait to be filmed: call film_scenes.` : ""}`, notes)
+      };
+    }
+    case "add_music": {
+      const run = needRun(ctx, "Music");
+      const prompt = str(args, "prompt", true);
+      const t = ctx.getTimeline();
+      const total = cutSec(t, sourceSec);
+      const at = Math.max(0, (_j = num(args, "at_sec")) != null ? _j : 0);
+      const wanted = (_k = num(args, "duration_sec")) != null ? _k : total > at ? total - at : DEFAULT_MUSIC_SEC;
+      const durationSec = Math.max(10, Math.min(MAX_MUSIC_SEC, Math.round(wanted)));
+      (_l = ctx.progress) == null ? void 0 : _l.call(ctx, `Composing music (${durationSec}s)\u2026`);
+      const audioUrl = await composeMusic(run, `Instrumental music bed, no vocals: ${prompt}`, durationSec);
+      const now = ctx.getTimeline();
+      if (at <= 0) {
+        ctx.setTimeline(__spreadProps(__spreadValues({}, now), {
+          soundtrack: {
+            audioUrl,
+            mode: "mix",
+            title: excerpt(prompt, 60)
+          }
+        }));
+        return {
+          ok: true,
+          label: "Generated music",
+          detail: excerpt(prompt, 60),
+          content: `Music laid under the whole cut (${secs(durationSec)}), mixed under the other audio.`
+        };
+      }
+      const audio = sceneAudioAt(now, {
+        audioUrl,
+        startSec: at,
+        durationSec: null,
+        volume: 0.6,
+        label: `Music \xB7 ${excerpt(prompt, 32)}`
+      }, sourceSec);
+      ctx.setTimeline(__spreadProps(__spreadValues({}, now), {
+        sceneAudio: [...(_m = now.sceneAudio) != null ? _m : [], audio]
+      }));
+      return {
+        ok: true,
+        label: "Generated music",
+        detail: excerpt(prompt, 60),
+        content: `Music placed at ${secs(at)} (id=${audio.id}).`
+      };
+    }
+    case "add_sound_effect": {
+      const run = needRun(ctx, "Sound effects");
+      const prompt = str(args, "prompt", true);
+      const at = Math.max(0, (_n = num(args, "at_sec")) != null ? _n : 0);
+      const durationSec = Math.max(0.5, Math.min(20, (_o = num(args, "duration_sec")) != null ? _o : DEFAULT_SFX_SEC));
+      (_p = ctx.progress) == null ? void 0 : _p.call(ctx, `Generating sound effect: ${excerpt(prompt, 40)}\u2026`);
+      const audioUrl = await composeSoundEffect(run, prompt, durationSec);
+      const now = ctx.getTimeline();
+      const audio = sceneAudioAt(now, {
+        audioUrl,
+        startSec: at,
+        durationSec,
+        volume: 0.9,
+        label: `SFX \xB7 ${excerpt(prompt, 32)}`
+      }, sourceSec);
+      ctx.setTimeline(__spreadProps(__spreadValues({}, now), {
+        sceneAudio: [...(_q = now.sceneAudio) != null ? _q : [], audio]
+      }));
+      return {
+        ok: true,
+        label: "Added sound effect",
+        detail: excerpt(prompt, 60),
+        content: `Sound effect placed at ${secs(at)} (id=${audio.id}).`
+      };
+    }
+    case "add_voiceover": {
+      const run = needRun(ctx, "Voiceover");
+      const text = str(args, "text", true);
+      const voice = str(args, "voice");
+      if (voice && !STOCK_VOICES.some((v) => v.id === voice)) throw new ToolInputError(`unknown voice "${voice}"`);
+      const before = ctx.getTimeline();
+      const lines = ((_r = before.sceneAudio) != null ? _r : []).filter((a) => {
+        var _a2;
+        return (_a2 = a.label) == null ? void 0 : _a2.startsWith("Voiceover");
+      });
+      const lastEnd = lines.reduce((end, a) => {
+        var _a2;
+        return Math.max(end, a.startSec + ((_a2 = a.durationSec) != null ? _a2 : 0));
+      }, 0);
+      const at = Math.max(0, (_s = num(args, "at_sec")) != null ? _s : lastEnd);
+      (_t = ctx.progress) == null ? void 0 : _t.call(ctx, "Recording voiceover\u2026");
+      const audioUrl = await speakLine(run, text, voice);
+      const measured = await probeAudioSec(run, audioUrl);
+      const now = ctx.getTimeline();
+      const audio = sceneAudioAt(now, {
+        audioUrl,
+        startSec: at,
+        durationSec: measured > 0 ? Math.round(measured * 1e3) / 1e3 : null,
+        volume: 1,
+        label: `Voiceover \xB7 ${excerpt(text, 32)}`
+      }, sourceSec);
+      ctx.setTimeline(__spreadProps(__spreadValues({}, now), {
+        sceneAudio: [...(_u = now.sceneAudio) != null ? _u : [], audio]
+      }));
+      return {
+        ok: true,
+        label: "Added voiceover",
+        detail: excerpt(text, 60),
+        content: `Voiceover placed at ${secs(at)} (id=${audio.id}${measured > 0 ? `, ${secs(measured)}` : ""}).`
+      };
+    }
+    case "add_captions": {
+      const ids = strList(args, "scene_ids");
+      const t = ctx.getTimeline();
+      const notes = [];
+      if (ids) {
+        for (const id of ids) if (!t.clips.some((c) => c.id === id)) notes.push(`no scene with id "${id}"`);
+      }
+      const scenes = t.clips.filter((c) => c.type === "video" && !c.fillColor && (!ids || ids.includes(c.id)));
+      let next = t;
+      let count = 0;
+      for (const scene of scenes) {
+        let cues = (_v = cache(ctx).get(scene.src)) == null ? void 0 : _v.transcript;
+        if (!cues && ctx.runAction) {
+          (_w = ctx.progress) == null ? void 0 : _w.call(ctx, `Transcribing ${scene.name}\u2026`);
+          cues = (await analyzeItem({
+            id: scene.id,
+            type: "video",
+            src: scene.src,
+            name: scene.name
+          }, ctx, false)).analysis.transcript;
+        }
+        const lines = captionsForClip(next, scene.id, cues != null ? cues : [], sourceSec);
+        count += lines.length;
+        next = placeCaptions(next, lines);
+      }
+      if (!count) notes.push("no dialogue in those scenes");
+      else ctx.setTimeline(next);
+      return {
+        ok: count > 0,
+        label: count > 0 ? `Captioned ${plural(count, "line")}` : "No captions added",
+        content: withNotes(`Added ${plural(count, "caption")}.`, notes)
+      };
+    }
+    case "set_format": {
+      const aspect = str(args, "aspect", true);
+      if (!(aspect in FORMAT_ASPECTS)) {
+        return {
+          ok: false,
+          label: "Format unchanged",
+          content: `Format unchanged.
+Notes: unknown aspect "${aspect}" - use 16:9, 9:16, 1:1, 4:5 or auto`
+        };
+      }
+      const t = ctx.getTimeline();
+      const ratio = FORMAT_ASPECTS[aspect];
+      const rest = __spreadValues({}, t);
+      delete rest.aspect;
+      ctx.setTimeline(ratio ? __spreadProps(__spreadValues({}, t), {
+        aspect: ratio
+      }) : rest);
+      return {
+        ok: true,
+        label: `Set format ${aspect}`,
+        content: `Format set to ${aspect}.`
+      };
+    }
+  }
+}export{CUT_AGENT_TOOLS,FORMAT_ASPECTS,MEDIA_TOOL_NAMES,SHOT_SEC_RANGE,agentLibrary,executeCutTool};//# sourceMappingURL=cutTools.js.map