@shop-os/modelkit 0.7.2 → 0.7.3

package/dist/index.js

@@ -13672,24 +13672,24 @@ import { z as z25 } from "zod";
 function mp4DurationSeconds(bytes) {
   const moov = findBox(bytes, 0, bytes.length, "moov");
   if (!moov) return void 0;
-  const mvhd = findBox(bytes, moov.start + 8, moov.end, "mvhd");
+  const mvhd = findBox(bytes, moov.body, moov.end, "mvhd");
   if (!mvhd) return void 0;
   const view = new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength);
-  const size = mvhd.end - mvhd.start;
-  if (size < 8 + 4) return void 0;
-  const version = view.getUint8(mvhd.start + 8);
+  const size = mvhd.end - mvhd.body;
+  if (size < 4) return void 0;
+  const version = view.getUint8(mvhd.body);
   if (version === 0) {
-    if (size < 8 + 4 + 4 + 4 + 4 + 4) return void 0;
-    const timescale = view.getUint32(mvhd.start + 8 + 4 + 4 + 4);
-    const duration = view.getUint32(mvhd.start + 8 + 4 + 4 + 4 + 4);
+    if (size < 4 + 4 + 4 + 4 + 4) return void 0;
+    const timescale = view.getUint32(mvhd.body + 4 + 4 + 4);
+    const duration = view.getUint32(mvhd.body + 4 + 4 + 4 + 4);
     if (timescale === 0) return void 0;
     return duration / timescale;
   }
   if (version === 1) {
-    if (size < 8 + 4 + 8 + 8 + 4 + 8) return void 0;
-    const timescale = view.getUint32(mvhd.start + 8 + 4 + 8 + 8);
-    const high = view.getUint32(mvhd.start + 8 + 4 + 8 + 8 + 4);
-    const low = view.getUint32(mvhd.start + 8 + 4 + 8 + 8 + 4 + 4);
+    if (size < 4 + 8 + 8 + 4 + 8) return void 0;
+    const timescale = view.getUint32(mvhd.body + 4 + 8 + 8);
+    const high = view.getUint32(mvhd.body + 4 + 8 + 8 + 4);
+    const low = view.getUint32(mvhd.body + 4 + 8 + 8 + 4 + 4);
     if (timescale === 0) return void 0;
     return (high * 2 ** 32 + low) / timescale;
   }
@@ -13699,15 +13699,23 @@ function findBox(bytes, from, to, type) {
   const view = new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength);
   let offset = from;
   while (offset + 8 <= to) {
-    const size = view.getUint32(offset);
+    let size = view.getUint32(offset);
+    let header = 8;
     const name = String.fromCharCode(
       view.getUint8(offset + 4),
       view.getUint8(offset + 5),
       view.getUint8(offset + 6),
       view.getUint8(offset + 7)
     );
-    if (size < 8 || offset + size > to) return void 0;
-    if (name === type) return { start: offset, end: offset + size };
+    if (size === 1) {
+      if (offset + 16 > to) return void 0;
+      size = view.getUint32(offset + 8) * 2 ** 32 + view.getUint32(offset + 12);
+      header = 16;
+    } else if (size === 0) {
+      size = to - offset;
+    }
+    if (size < header || offset + size > to) return void 0;
+    if (name === type) return { body: offset + header, end: offset + size };
     offset += size;
   }
   return void 0;
@@ -13720,6 +13728,7 @@ var IMAGE_TIMEOUT_MS = 5 * 6e4;
 var VIDEO_POLL_INTERVAL_MS = 5e3;
 var VIDEO_TIMEOUT_MS = 20 * 6e4;
 var PROMPT_MAX_CHARS = 6e3;
+var OUTPUT_URL_TTL_MS = 55 * 6e4;
 var LUMA_IMAGE_MODEL_IDS = /* @__PURE__ */ new Set(["uni-1", "uni-1-max"]);
 var LUMA_VIDEO_MODEL_ID = "ray-3.2";
 var LUMA_VIDEO_MODES = /* @__PURE__ */ new Map([
@@ -13738,6 +13747,9 @@ function lumaRowRefusal(row2) {
   }
   return `luma has no ${row2.type} rows`;
 }
+function lumaModel(ctx) {
+  return ctx.model.providerModelId.toLowerCase();
+}
 function lumaBase(ctx) {
   return ctx.endpoints?.luma ?? DEFAULT_BASE_URL;
 }
@@ -13780,8 +13792,23 @@ function imageRef(ref) {
   if ("url" in ref) return { url: ref.url };
   return { data: mediaRefToBase64(ref.bytes), media_type: ref.mime };
 }
+var VIDEO_TYPES_BY_EXTENSION = /* @__PURE__ */ new Map([
+  ["mp4", "video/mp4"],
+  ["m4v", "video/mp4"],
+  ["mov", "video/quicktime"],
+  ["webm", "video/webm"]
+]);
+function videoUrlType(url) {
+  let path = url;
+  try {
+    path = new URL(url).pathname;
+  } catch {
+  }
+  const extension = path.slice(path.lastIndexOf(".") + 1).toLowerCase();
+  return VIDEO_TYPES_BY_EXTENSION.get(extension) ?? "video/mp4";
+}
 function videoRef(ref) {
-  if ("url" in ref) return { url: ref.url, media_type: "video/mp4" };
+  if ("url" in ref) return { url: ref.url, media_type: videoUrlType(ref.url) };
   const mediaType = ref.mime.startsWith("video/") ? ref.mime : "video/mp4";
   return { data: mediaRefToBase64(ref.bytes), media_type: mediaType };
 }
@@ -13945,14 +13972,22 @@ async function lumaRequest(ctx, url, body) {
   }
   return res;
 }
+async function parseGeneration(res) {
+  try {
+    const parsed = lumaGenerationSchema.safeParse(await res.json());
+    return parsed.success ? parsed.data : void 0;
+  } catch {
+    return "not_json";
+  }
+}
 async function submitGeneration(ctx, body) {
   const retryOptions = mediaRetryOptions(ctx.call, ctx.signal);
   const res = await withRetry(
     () => lumaRequest(ctx, `${lumaBase(ctx)}/v1/generations`, JSON.stringify(body)),
     retryOptions
   );
-  const parsed = lumaGenerationSchema.safeParse(await res.json());
-  if (!parsed.success) {
+  const generation = await parseGeneration(res);
+  if (generation === void 0 || generation === "not_json") {
     throw new ModelKitError({
       code: "provider_rejected",
       message: "Luma submit returned no generation id",
@@ -13961,13 +13996,20 @@ async function submitGeneration(ctx, body) {
       retryable: false
     });
   }
-  return parsed.data;
+  return generation;
 }
 async function readGeneration(ctx, id) {
   const res = await lumaRequest(ctx, `${lumaBase(ctx)}/v1/generations/${id}`);
-  const payload = JsonValueSchema.parse(await res.json());
-  const parsed = lumaGenerationSchema.safeParse(payload);
-  if (!parsed.success) {
+  const generation = await parseGeneration(res);
+  if (generation === "not_json") {
+    throw new ModelKitError({
+      code: "network",
+      message: "Luma returned a poll response that is not JSON",
+      provider: "luma",
+      model: ctx.model.id
+    });
+  }
+  if (generation === void 0) {
     throw new ModelKitError({
       code: "provider_rejected",
       message: "Luma returned an unreadable generation",
@@ -13976,14 +14018,12 @@ async function readGeneration(ctx, id) {
       retryable: false
     });
   }
-  return parsed.data;
+  return generation;
 }
 function outputUrl(generation, type) {
   const outputs = generation.output ?? [];
-  const match = outputs.find((item) => item.type === type && item.url);
-  if (match?.url) return match.url;
-  const fallback = outputs.find((item) => item.url);
-  return fallback?.url;
+  const match = outputs.find((item) => item.type === type && item.url) ?? outputs.find((item) => item.type === void 0 && item.url);
+  return match?.url;
 }
 async function readGenerationOnce(ctx, id) {
   try {
@@ -14022,21 +14062,22 @@ async function checkGeneration(ctx, id, type) {
   return void 0;
 }
 async function downloadBytes(ctx, url, fallback) {
-  let bytes;
-  let header;
-  try {
-    const out = await mediaRefToBytes({ url }, ctx.fetch, {
-      ...ctx.signal !== void 0 && { signal: ctx.signal },
-      provider: "luma",
-      model: ctx.model.id
-    });
-    bytes = out.bytes;
-    header = out.mime;
-  } catch (cause) {
-    throw classifyThrown("luma", cause, ctx.model.id);
-  }
+  const { bytes, mime: header } = await withRetry(
+    async () => {
+      try {
+        return await mediaRefToBytes({ url }, ctx.fetch, {
+          ...ctx.signal !== void 0 && { signal: ctx.signal },
+          provider: "luma",
+          model: ctx.model.id
+        });
+      } catch (cause) {
+        throw classifyThrown("luma", cause, ctx.model.id);
+      }
+    },
+    mediaRetryOptions(ctx.call, ctx.signal)
+  );
   const kind = fallback.split("/")[0];
-  const usable = header !== void 0 && header.startsWith(`${kind}/`);
+  const usable = header.startsWith(`${kind}/`);
   return { bytes, mime: usable ? header : fallback };
 }
 function imageUsage2(ctx) {
@@ -14051,15 +14092,12 @@ function imageUsage2(ctx) {
 }
 var GENERATE_SECONDS = 5;
 var MAX_PLAUSIBLE_SECONDS = 60;
-function billedSeconds(req, mode, bytes) {
-  if (mode === "generate") return GENERATE_SECONDS;
+function editedSeconds(req, bytes) {
   const measured = mp4DurationSeconds(bytes);
-  const source2 = req.sourceVideo;
-  const length = measured !== void 0 && measured > 0 && measured <= MAX_PLAUSIBLE_SECONDS ? measured : source2 !== void 0 && "url" in source2 ? source2.durationSeconds : void 0;
+  const length = measured !== void 0 && measured > 0 && measured <= MAX_PLAUSIBLE_SECONDS ? measured : req.sourceVideo?.durationSeconds;
   return Math.max(length ?? 0, GENERATE_SECONDS);
 }
-function videoUsage4(ctx, req, mode, bytes) {
-  const seconds3 = billedSeconds(req, mode, bytes);
+function videoUsage4(ctx, req, mode, seconds3) {
   const event = {
     model: ctx.model.id,
     provider: "luma",
@@ -14074,7 +14112,7 @@ function videoUsage4(ctx, req, mode, bytes) {
 }
 function imageBody(req, ctx, outputFormat) {
   const prompt = checkPrompt(req, ctx);
-  const model = ctx.model.providerModelId;
+  const model = lumaModel(ctx);
   const ratio = req.aspectRatio !== "adaptive" ? req.aspectRatio : void 0;
   const [source2, ...rest] = req.references ?? [];
   if (source2 === void 0) {
@@ -14102,7 +14140,7 @@ function imageBody(req, ctx, outputFormat) {
 }
 function videoBody(req, ctx, mode) {
   const prompt = checkPrompt(req, ctx);
-  const model = ctx.model.providerModelId;
+  const model = lumaModel(ctx);
   const hasReferences = (req.referenceImages ?? []).length + (req.referenceVideos ?? []).length + (req.referenceAudios ?? []).length > 0;
   if (hasReferences) {
     refuseOrDrop(
@@ -14184,7 +14222,7 @@ function videoBody(req, ctx, mode) {
 async function generateLumaImage(req, ctx) {
   throwIfAborted(ctx.signal);
   lumaKey(ctx);
-  if (!LUMA_IMAGE_MODEL_IDS.has(ctx.model.providerModelId.toLowerCase())) {
+  if (!LUMA_IMAGE_MODEL_IDS.has(lumaModel(ctx))) {
     throw new ModelKitError({
       code: "unsupported_option",
       message: `luma has no image mapping for "${ctx.model.providerModelId}"`,
@@ -14217,16 +14255,25 @@ function lumaVideoJob(ctx, req, mode, id) {
       return { status: "pending" };
     },
     async result(opts) {
-      const signal = opts?.signal ?? ctx.signal;
-      const url = await pollUntil(checkGeneration.bind(null, ctx, id, "video"), {
+      const signal = combineSignals(ctx.signal, opts?.signal);
+      const jobCtx = { ...ctx, ...signal !== void 0 && { signal } };
+      const url = await pollUntil(checkGeneration.bind(null, jobCtx, id, "video"), {
         intervalMs: opts?.pollIntervalMs ?? VIDEO_POLL_INTERVAL_MS,
         timeoutMs: opts?.timeoutMs ?? VIDEO_TIMEOUT_MS,
         ...signal !== void 0 && { signal }
       });
-      const video = await downloadBytes(ctx, url, "video/mp4");
+      if (mode === "generate") {
+        const result2 = {
+          video: { url, expiresAt: new Date(Date.now() + OUTPUT_URL_TTL_MS) },
+          usage: videoUsage4(jobCtx, req, mode, GENERATE_SECONDS),
+          raw: { id }
+        };
+        return result2;
+      }
+      const video = await downloadBytes(jobCtx, url, "video/mp4");
       const result = {
         video,
-        usage: videoUsage4(ctx, req, mode, video.bytes),
+        usage: videoUsage4(jobCtx, req, mode, editedSeconds(req, video.bytes)),
         raw: { id }
       };
       return result;
@@ -14236,7 +14283,7 @@ function lumaVideoJob(ctx, req, mode, id) {
 async function generateLumaVideo(req, ctx) {
   throwIfAborted(ctx.signal);
   lumaKey(ctx);
-  if (ctx.model.providerModelId.toLowerCase() !== LUMA_VIDEO_MODEL_ID) {
+  if (lumaModel(ctx) !== LUMA_VIDEO_MODEL_ID) {
     throw new ModelKitError({
       code: "unsupported_option",
       message: `luma has no video mapping for "${ctx.model.providerModelId}"`,