@koda-sl/baker-cli 0.306.3-dev.bb28562ac → 0.306.6

package/dist/cli.js

@@ -67,7 +67,7 @@ import {
   validateCanvasDeep,
   withoutCastName,
   ytDlpBlockSignal
-} from "./chunk-3TEVIAAV.js";
+} from "./chunk-WONWYE7B.js";
 import {
   csvOrJson,
   daysAgoIso,
@@ -76,7 +76,7 @@ import {
   resolveAccountIdArg,
   resolveEffectiveStatus,
   todayIso
-} from "./chunk-4YF56EGK.js";
+} from "./chunk-KMQHIAYG.js";
 import {
   buildQueryCacheKey,
   cacheGet,
@@ -92,32 +92,32 @@ import {
   writeAdsJson,
   writeAdsOutput,
   writeJsonEnvelope
-} from "./chunk-A2VZOTCV.js";
+} from "./chunk-AEU5X443.js";
 import {
   ApiError,
   apiGet,
   apiGetWithHeaders,
   apiPost,
   validateConvexId
-} from "./chunk-KDTHRRAC.js";
+} from "./chunk-TTIGSGXE.js";
 import {
   installStreamTaps,
   logInvocation
-} from "./chunk-2IUN3S3E.js";
+} from "./chunk-FOWT5AS4.js";
 import {
   captureBudgetMs,
   captureProxyCredentials,
   captureVimeoCookie,
-  frameVisionKey,
   getEnv,
-  hasFrameVisionKey,
+  hasOpenRouterKey,
   isProxyFailure,
+  openRouterKey,
   plannedRoutes,
   refuseNonPublicUrl,
   requireChatId,
   resolveChatId,
   shouldEscalate
-} from "./chunk-WFWU3CHS.js";
+} from "./chunk-UIQJZGBG.js";
 
 // src/cli.ts
 import { defineCommand as defineCommand240, runMain } from "citty";
@@ -23377,10 +23377,10 @@ Example: baker ads meta ads retag 120252158038400247 --url-tags "utm_source={{si
     }
   },
   run: async ({ args }) => {
-    const { apiPost: apiPost2 } = await import("./client-6KQHCXS2.js");
-    const { requireChatId: requireChatId2 } = await import("./env-C7YXL6C6.js");
-    const { writeJsonEnvelope: writeJsonEnvelope2 } = await import("./output-RBM32FKJ.js");
-    const { handleMetaError: handleMetaError2 } = await import("./shared-WJIJTWST.js");
+    const { apiPost: apiPost2 } = await import("./client-QFJHI3FX.js");
+    const { requireChatId: requireChatId2 } = await import("./env-PJSBLG56.js");
+    const { writeJsonEnvelope: writeJsonEnvelope2 } = await import("./output-PAE5S7RI.js");
+    const { handleMetaError: handleMetaError2 } = await import("./shared-J6GGQDIL.js");
     try {
       const accountId = bareAccountId2(args);
       const urlTags = typeof args["url-tags"] === "string" ? args["url-tags"] : void 0;
@@ -23521,10 +23521,10 @@ function duplicateCommand2(entity, label) {
       replace: { type: "boolean", description: "Pause the original once the copy publishes" }
     },
     run: async ({ args }) => {
-      const { apiPost: apiPost2 } = await import("./client-6KQHCXS2.js");
-      const { requireChatId: requireChatId2 } = await import("./env-C7YXL6C6.js");
-      const { writeJsonEnvelope: writeJsonEnvelope2 } = await import("./output-RBM32FKJ.js");
-      const { handleMetaError: handleMetaError2 } = await import("./shared-WJIJTWST.js");
+      const { apiPost: apiPost2 } = await import("./client-QFJHI3FX.js");
+      const { requireChatId: requireChatId2 } = await import("./env-PJSBLG56.js");
+      const { writeJsonEnvelope: writeJsonEnvelope2 } = await import("./output-PAE5S7RI.js");
+      const { handleMetaError: handleMetaError2 } = await import("./shared-J6GGQDIL.js");
       try {
         const accountId = bareAccountId2(args);
         const chatId = requireChatId2();
@@ -30402,7 +30402,7 @@ function rosterHints(roster, statusFilter) {
     `@${mine[0]} was made in this conversation${mine.length > 1 ? ` (with ${mine.slice(1).map((h) => `@${h}`).join(", ")})` : ""} \u2014 that is the avatar this chat is about, and the user chose its face. Use it. The others belong to earlier conversations: only reach for one if the user names it, and never ask which to use when the answer is the one you just cast.`
   ] : [];
   hints2.push(
-    "Casting one of these? `baker avatars get <handle>` returns the subject description you copy into the prompt verbatim plus the identity sheet URL for `--reference`. The roster leaves the description out on purpose \u2014 it must never be read half-way."
+    "Casting one of these? Cast them with `--avatar <handle>` on `baker studio generate` / `baker studio animate` \u2014 that flag carries their face AND their accent. `baker avatars get <handle>` returns the written profile behind it; the roster leaves the description out on purpose, because it must never be read half-way."
   );
   const building = roster.filter((avatar) => avatar.status === "generating").map((avatar) => avatar.handle);
   if (building.length > 0) {
@@ -36920,17 +36920,20 @@ import { tmpdir } from "os";
 import path15 from "path";
 import { promisify as promisify2 } from "util";
 import { z as z35 } from "zod";
-var RENDER_REVIEW_MODEL = "gemini-pro-latest";
+var RENDER_REVIEW_MODEL = "~google/gemini-pro-latest";
 var RENDER_REVIEW_FRAMES = 12;
 var RENDER_REVIEW_WIDTH = 1024;
-var RENDER_WATCH_MODEL = "gemini-3.7-flash";
+var RENDER_WATCH_MODEL = "~google/gemini-flash-latest";
 function usageFrom(model, body) {
-  const meta = body?.usageMetadata;
-  if (!meta) return null;
-  const inputTokens = meta.promptTokenCount ?? 0;
-  const outputTokens = meta.candidatesTokenCount ?? 0;
-  if (inputTokens === 0 && outputTokens === 0) return null;
-  return { model, inputTokens, outputTokens };
+  const usage = body?.usage;
+  if (!usage) return null;
+  const positive = (value) => typeof value === "number" && Number.isFinite(value) && value > 0 ? value : null;
+  return {
+    model,
+    inputTokens: usage.prompt_tokens ?? 0,
+    outputTokens: usage.completion_tokens ?? 0,
+    costUsd: positive(usage.cost) ?? positive(usage.cost_details?.upstream_inference_cost) ?? 0
+  };
 }
 var WATCH_MAX_BYTES = 60 * 1024 * 1024;
 function sceneAt(windows, atS) {
@@ -37233,7 +37236,7 @@ function classifyAudio(volumedetect, hasAudioStream) {
   }
   return findings;
 }
-var RENDER_REVIEW_UNAVAILABLE = "The render was NOT reviewed \u2014 no frame-vision key in this environment. Watch it yourself before sending it on.";
+var RENDER_REVIEW_UNAVAILABLE = "The render was NOT reviewed \u2014 no model gateway key in this environment. Watch it yourself before sending it on.";
 var REVIEW_CONCURRENCY = 3;
 async function mapWithLimit(items, limit, fn) {
   const out = new Array(items.length);
@@ -37276,64 +37279,63 @@ async function videoDuration(video) {
     return null;
   }
 }
-async function askVideo(video, prompt, spent) {
-  const key = frameVisionKey();
+async function askGateway(model, parts, timeoutMs, spent) {
+  const key = openRouterKey();
   if (!key) return null;
   try {
-    const bytes = await readFile11(video);
-    if (bytes.byteLength > WATCH_MAX_BYTES) return null;
-    const response = await fetch(
-      `https://generativelanguage.googleapis.com/v1beta/models/${RENDER_WATCH_MODEL}:generateContent`,
-      {
-        method: "POST",
-        headers: { "content-type": "application/json", "x-goog-api-key": key },
-        body: JSON.stringify({
-          contents: [
-            {
-              parts: [{ text: prompt }, { inline_data: { mime_type: "video/mp4", data: bytes.toString("base64") } }]
-            }
-          ]
-        }),
-        signal: AbortSignal.timeout(6e5)
-      }
-    );
+    const response = await fetch("https://openrouter.ai/api/v1/chat/completions", {
+      method: "POST",
+      headers: { "content-type": "application/json", authorization: `Bearer ${key}` },
+      body: JSON.stringify({
+        model,
+        messages: [{ role: "user", content: parts }],
+        usage: { include: true }
+      }),
+      signal: AbortSignal.timeout(timeoutMs)
+    });
     if (!response.ok) return null;
     const body = await response.json();
-    const usage = usageFrom(RENDER_WATCH_MODEL, body);
+    const usage = usageFrom(model, body);
     if (usage) spent(usage);
-    const text2 = body.candidates?.[0]?.content?.parts?.map((p) => p.text ?? "").join("");
+    const text2 = body.choices?.[0]?.message?.content;
     return text2 ? parseReviewJson(text2) : null;
   } catch {
     return null;
   }
 }
-async function ask(images, prompt, spent) {
-  const key = frameVisionKey();
-  if (!key) return null;
-  const parts = [{ text: prompt }];
-  for (const b64 of images) parts.push({ inline_data: { mime_type: "image/jpeg", data: b64 } });
+async function askVideo(video, prompt, spent) {
+  let bytes;
   try {
-    const response = await fetch(
-      `https://generativelanguage.googleapis.com/v1beta/models/${RENDER_REVIEW_MODEL}:generateContent`,
-      {
-        method: "POST",
-        headers: { "content-type": "application/json", "x-goog-api-key": key },
-        body: JSON.stringify({ contents: [{ parts }] }),
-        signal: AbortSignal.timeout(18e4)
-      }
-    );
-    if (!response.ok) return null;
-    const body = await response.json();
-    const usage = usageFrom(RENDER_REVIEW_MODEL, body);
-    if (usage) spent(usage);
-    const text2 = body.candidates?.[0]?.content?.parts?.map((p) => p.text ?? "").join("");
-    return text2 ? parseReviewJson(text2) : null;
+    bytes = await readFile11(video);
   } catch {
     return null;
   }
+  if (bytes.byteLength > WATCH_MAX_BYTES) return null;
+  return await askGateway(
+    RENDER_WATCH_MODEL,
+    [
+      { type: "text", text: prompt },
+      { type: "video_url", video_url: { url: `data:video/mp4;base64,${bytes.toString("base64")}` } }
+    ],
+    6e5,
+    spent
+  );
+}
+async function ask(images, prompt, spent) {
+  return await askGateway(
+    RENDER_REVIEW_MODEL,
+    [
+      { type: "text", text: prompt },
+      ...images.map(
+        (b64) => ({ type: "image_url", image_url: { url: `data:image/jpeg;base64,${b64}` } })
+      )
+    ],
+    18e4,
+    spent
+  );
 }
 async function reviewRenderedVideo(opts) {
-  if (!hasFrameVisionKey()) return null;
+  if (!hasOpenRouterKey()) return null;
   const duration = await videoDuration(opts.video);
   if (duration === null) return null;
   const spent = (usage) => opts.onSpend?.(usage);
@@ -37815,6 +37817,12 @@ ${describeRewrites(healed.rewrites)}
     }
     if (e instanceof RunAbortedError && e.reason === "checkpoint") {
       await clearRunMarker(outputsDir, filePath);
+      if (poster) {
+        await poster.flush({
+          ...failedPayload(`Stopped before the clips, waiting on a decision. Resume with --run-id ${runId}.`),
+          interrupted: true
+        });
+      }
       const frames = frameUrls;
       process.stdout.write(
         `${JSON.stringify(
@@ -39335,6 +39343,13 @@ function beatToCamera(spec, beat, isClosingCard) {
   if (beat.on_camera !== void 0) return beat.on_camera;
   return Boolean(spec.cast) && beat.cast !== false;
 }
+function beatsIntroducingSpeakerOffCamera(spec) {
+  const closesOnCard = endCardWanted(spec);
+  const last = spec.beats.length - 1;
+  return spec.beats.flatMap(
+    (beat, i) => lineIntroducesSpeaker(beat.say) && !beatToCamera(spec, beat, closesOnCard && i === last) ? [i] : []
+  );
+}
 function wantsNativeSpeech(spec, closesOnCard) {
   if (!spec.cast || spec.voiceover === false) return false;
   const last = spec.beats.length - 1;
@@ -39453,7 +39468,7 @@ function beatsTooShortToPlay(spec) {
 }
 
 // src/engine/scaffold/ad-brand-overlay.ts
-var markFallback = (name) => name.trim() ? ` onerror="this.replaceWith(Object.assign(document.createElement('span'),{className:this.className+' brand-wordmark',textContent:${JSON.stringify(name.trim())}}))"` : "";
+var markFallback = (name) => name.trim() ? ` onerror="this.replaceWith(Object.assign(document.createElement('span'),{className:this.className+' brand-wordmark',textContent:${escapeHtml2(JSON.stringify(name.trim()))}}))"` : "";
 var escapeHtml2 = (s) => s.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;").replace(/"/g, "&quot;");
 function firstHex(palette) {
   const hex = (palette ?? []).find((h) => /^#?[0-9a-fA-F]{6}$/.test(h.trim()));
@@ -39822,7 +39837,12 @@ var scaffoldAdCommand = defineCommand111({
         error: {
           code: "VALIDATION_ERROR",
           message: "This ad names a recurring place but gives no picture of it. A location sheet is fused from a real image; with nothing to fuse there is no sheet, and every shot invents the room again.",
-          fix: `Make the plate once, then pass it: \`baker images generate "${spec.data.set.description}" --model openai/gpt-image-2\`, then put its url in the spec as \`"set": { "description": "\u2026", "image": "<url>" }\`. Drop \`set\` entirely if the ad does not keep returning to one place.`
+          // Names `studio generate`, which is the command that exists — `baker images`
+          // finds and ingests pictures, it does not make them, so the line this used to
+          // print could not be run. No `--model` either: the lane already defaults to the
+          // curated image model, and the one named here was the legacy id the picker now
+          // offers to nobody.
+          fix: `Make the plate once, then pass it: \`baker studio generate "${spec.data.set.description}"\`, then put its url in the spec as \`"set": { "description": "\u2026", "image": "<url>" }\`. Drop \`set\` entirely if the ad does not keep returning to one place.`
         }
       });
       process.exit(1);
@@ -39884,14 +39904,15 @@ var scaffoldAdCommand = defineCommand111({
         ...filledPalette ? { palette: filledPalette } : {}
       };
     }
-    const introducesOffCamera = spec.data.beats.map((beat, i) => ({ i, beat })).filter(({ beat }) => beat.on_camera !== true && lineIntroducesSpeaker(beat.say));
-    if (introducesOffCamera.length > 0) {
-      const first = introducesOffCamera[0];
+    const introducesOffCamera = beatsIntroducingSpeakerOffCamera(spec.data);
+    const firstIntro = introducesOffCamera[0];
+    if (firstIntro !== void 0) {
+      const line = (spec.data.beats[firstIntro]?.say ?? "").trim();
       writeJson({
         ok: false,
         error: {
           code: "VALIDATION_ERROR",
-          message: `Beat ${(first?.i ?? 0) + 1} says "${first?.beat.say.trim()}" \u2014 the speaker introducing themselves \u2014 but nobody is on camera saying it. A voice from off screen introducing the person standing in shot reads as a mistake, the same way a moving mouth under someone else's voice does.`,
+          message: `Beat ${firstIntro + 1} says "${line}" \u2014 the speaker introducing themselves \u2014 but nobody is on camera saying it. A voice from off screen introducing the person standing in shot reads as a mistake, the same way a moving mouth under someone else's voice does.`,
           fix: 'Either set `on_camera: true` on that beat so they say it themselves \u2014 which needs the avatar to have their own accent \u2014 or rewrite the line as narration about them rather than by them: "Marta instal\xF3 placas en nueve d\xEDas", not "Hola, soy Marta".'
         }
       });