@shop-os/modelkit 0.4.5 → 0.4.7

package/dist/index.cjs

@@ -52,6 +52,7 @@ __export(index_exports, {
   SurfaceSchema: () => SurfaceSchema,
   VIDEO_INPUT_SPEC_KEYS: () => VIDEO_INPUT_SPEC_KEYS,
   VIDEO_TOOL_MODELS: () => VIDEO_TOOL_MODELS,
+  VideoResolutionSchema: () => VideoResolutionSchema,
   adapterServesType: () => adapterServesType,
   aiSdkTextAdapter: () => aiSdkTextAdapter,
   arkAdapter: () => arkAdapter,
@@ -3770,6 +3771,11 @@ var arkAdapter = {
   async generateVideo(req, ctx) {
     throwIfAborted(ctx.signal);
     return generateSeedanceVideo(req, ctx);
+  },
+  resumeVideo(req, jobId, ctx) {
+    throwIfAborted(ctx.signal);
+    const key = arkKey(ctx);
+    return seedanceJob(ctx, key, jobId, seedanceUsage(ctx, req, seedanceInitialDuration(req)));
   }
 };
 function arkKey(ctx) {
@@ -3927,17 +3933,22 @@ var SeedanceResponseSchema = import_zod7.z.object({
   error: import_zod7.z.union([import_zod7.z.string(), import_zod7.z.object({ message: import_zod7.z.string().optional() })]).optional()
 });
 var SeedanceErrorMessageSchema = import_zod7.z.object({ message: import_zod7.z.string().optional() });
+function seedanceResolution(req) {
+  return req.resolution ?? "720p";
+}
 function seedanceUsage(ctx, req, duration) {
-  const usage2 = {
+  return {
     model: ctx.model.id,
     provider: "ark",
     modality: "VIDEO",
     unit: "seconds",
     quantity: duration,
-    durationSeconds: duration
+    durationSeconds: duration,
+    resolution: seedanceResolution(req)
   };
-  if (req.resolution !== void 0) usage2.resolution = req.resolution;
-  return usage2;
+}
+function seedanceInitialDuration(req) {
+  return req.durationSeconds ?? 5;
 }
 function seedanceErrorMessage(error, status) {
   const asString = import_zod7.z.string().safeParse(error);
@@ -3946,6 +3957,82 @@ function seedanceErrorMessage(error, status) {
   if (asObject.success && asObject.data.message !== void 0) return asObject.data.message;
   return `Seedance task ${status}`;
 }
+function seedanceJob(ctx, key, jobId, usage2) {
+  const modelId = ctx.model.id;
+  const fetchStatus = async () => {
+    const res = await arkFetch(ctx, `${arkBaseUrl(ctx)}${SUBMIT_PATH}/${jobId}`, {
+      headers: { Authorization: `Bearer ${key}` }
+    });
+    if (!res.ok) {
+      if (ctx.call.maxRetries !== 0 && (res.status === 429 || res.status >= 500))
+        return void 0;
+      const text2 = await res.text().catch(() => "");
+      throwArkError(res.status, text2, modelId);
+    }
+    const parsed = SeedanceResponseSchema.safeParse(await res.json());
+    if (!parsed.success) return {};
+    return parsed.data;
+  };
+  return {
+    jobId,
+    async poll() {
+      const polled = await fetchStatus();
+      if (!polled) return { status: "running" };
+      const status = (polled.status ?? "queued").toLowerCase();
+      if (status === "succeeded") return { status: "succeeded", raw: polled };
+      if (status === "failed" || status === "expired" || status === "cancelled") {
+        return { status: "failed", raw: polled };
+      }
+      if (status === "queued" || status === "pending")
+        return { status: "pending", raw: polled };
+      return { status: "running", raw: polled };
+    },
+    async result(opts) {
+      const optSignal = opts?.signal ?? ctx.signal;
+      const pollOpts = {
+        intervalMs: opts?.pollIntervalMs ?? SEEDANCE_POLL_INTERVAL_MS,
+        timeoutMs: opts?.timeoutMs ?? RESULT_TIMEOUT_MS
+      };
+      if (optSignal !== void 0) pollOpts.signal = optSignal;
+      return pollUntil(async () => {
+        const polled = await fetchStatus();
+        if (!polled) return void 0;
+        const status = (polled.status ?? "queued").toLowerCase();
+        if (status === "succeeded") {
+          const url = polled.content?.video_url;
+          if (!url)
+            throwArkError(
+              void 0,
+              "Seedance task succeeded but returned no video_url",
+              modelId
+            );
+          if (polled.duration !== void 0 && polled.duration > 0) {
+            usage2.quantity = polled.duration;
+            usage2.durationSeconds = polled.duration;
+          }
+          const rendered = VideoResolutionSchema.safeParse(
+            polled.resolution?.toLowerCase()
+          );
+          if (rendered.success) usage2.resolution = rendered.data;
+          const tokens = polled.usage;
+          if (tokens?.total_tokens !== void 0 && tokens.completion_tokens !== void 0) {
+            usage2.inputTokens = tokens.total_tokens - tokens.completion_tokens;
+            usage2.outputTokens = tokens.completion_tokens;
+          }
+          return {
+            video: { url, expiresAt: new Date(Date.now() + VIDEO_URL_TTL_MS) },
+            usage: usage2,
+            raw: polled
+          };
+        }
+        if (status === "failed" || status === "expired" || status === "cancelled") {
+          throwArkError(void 0, seedanceErrorMessage(polled.error, status), modelId);
+        }
+        return void 0;
+      }, pollOpts);
+    }
+  };
+}
 async function generateSeedanceVideo(req, ctx) {
   const modelId = ctx.model.id;
   const taskType = req.taskType;
@@ -3960,10 +4047,10 @@ async function generateSeedanceVideo(req, ctx) {
     });
   }
   const key = arkKey(ctx);
-  const requestedDuration = req.durationSeconds ?? 5;
+  const requestedDuration = seedanceInitialDuration(req);
   let duration = requestedDuration;
   if (taskType === "edit") duration = -1;
-  const resolution = req.resolution ?? "720p";
+  const resolution = seedanceResolution(req);
   const content = [{ type: "text", text: req.prompt }];
   if (req.firstFrame) {
     content.push({
@@ -4041,75 +4128,7 @@ async function generateSeedanceVideo(req, ctx) {
   if (!submit.id) throwArkError(void 0, "Seedance task create returned no id", modelId);
   const jobId = submit.id;
   const usage2 = seedanceUsage(ctx, req, taskType === "edit" ? requestedDuration : duration);
-  const fetchStatus = async () => {
-    const res = await arkFetch(ctx, `${arkBaseUrl(ctx)}${SUBMIT_PATH}/${jobId}`, {
-      headers: { Authorization: `Bearer ${key}` }
-    });
-    if (!res.ok) {
-      if (ctx.call.maxRetries !== 0 && (res.status === 429 || res.status >= 500))
-        return void 0;
-      const text2 = await res.text().catch(() => "");
-      throwArkError(res.status, text2, modelId);
-    }
-    const parsed = SeedanceResponseSchema.safeParse(await res.json());
-    if (!parsed.success) return {};
-    return parsed.data;
-  };
-  return {
-    jobId,
-    async poll() {
-      const polled = await fetchStatus();
-      if (!polled) return { status: "running" };
-      const status = (polled.status ?? "queued").toLowerCase();
-      if (status === "succeeded") return { status: "succeeded", raw: polled };
-      if (status === "failed" || status === "expired" || status === "cancelled") {
-        return { status: "failed", raw: polled };
-      }
-      if (status === "queued" || status === "pending")
-        return { status: "pending", raw: polled };
-      return { status: "running", raw: polled };
-    },
-    async result(opts) {
-      const optSignal = opts?.signal ?? ctx.signal;
-      const pollOpts = {
-        intervalMs: opts?.pollIntervalMs ?? SEEDANCE_POLL_INTERVAL_MS,
-        timeoutMs: opts?.timeoutMs ?? RESULT_TIMEOUT_MS
-      };
-      if (optSignal !== void 0) pollOpts.signal = optSignal;
-      return pollUntil(async () => {
-        const polled = await fetchStatus();
-        if (!polled) return void 0;
-        const status = (polled.status ?? "queued").toLowerCase();
-        if (status === "succeeded") {
-          const url = polled.content?.video_url;
-          if (!url)
-            throwArkError(
-              void 0,
-              "Seedance task succeeded but returned no video_url",
-              modelId
-            );
-          if (polled.duration !== void 0 && polled.duration > 0) {
-            usage2.quantity = polled.duration;
-            usage2.durationSeconds = polled.duration;
-          }
-          const tokens = polled.usage;
-          if (tokens?.total_tokens !== void 0 && tokens.completion_tokens !== void 0) {
-            usage2.inputTokens = tokens.total_tokens - tokens.completion_tokens;
-            usage2.outputTokens = tokens.completion_tokens;
-          }
-          return {
-            video: { url, expiresAt: new Date(Date.now() + VIDEO_URL_TTL_MS) },
-            usage: usage2,
-            raw: polled
-          };
-        }
-        if (status === "failed" || status === "expired" || status === "cancelled") {
-          throwArkError(void 0, seedanceErrorMessage(polled.error, status), modelId);
-        }
-        return void 0;
-      }, pollOpts);
-    }
-  };
+  return seedanceJob(ctx, key, jobId, usage2);
 }
 
 // src/adapters/fal.ts
@@ -4899,9 +4918,15 @@ function videoUsage(ctx, req, duration) {
     quantity: duration,
     durationSeconds: duration
   };
-  if (req.resolution !== void 0) event.resolution = req.resolution;
+  const resolution = falVideoResolution(ctx.model.providerModelId, req);
+  if (resolution !== void 0) event.resolution = resolution;
   return event;
 }
+function falVideoResolution(providerModelId, req) {
+  if (isWan(providerModelId)) return wanResolution(req);
+  if (isMinimax(providerModelId)) return minimaxResolution(req, minimaxFamily(providerModelId));
+  return req.resolution;
+}
 async function decodeImage(ctx, url, contentType) {
   if (url.startsWith("data:")) {
     const comma = url.indexOf(",");
@@ -5033,12 +5058,14 @@ function isFalVideoKnown(providerModelId) {
 function minimaxFamily(providerModelId) {
   return H3_MAX_MODEL_IDS.has(providerModelId.toLowerCase()) ? "h3-max" : "h3";
 }
+function minimaxResolution(req, family) {
+  return req.resolution ?? (family === "h3-max" ? "768p" : "2k");
+}
 function minimaxBase(req, family) {
-  const resolution = (family === "h3-max" ? req.resolution ?? "768p" : req.resolution ?? "2k").toUpperCase();
   const input = {
     prompt: req.prompt.trim(),
     duration: req.durationSeconds ?? 5,
-    resolution,
+    resolution: minimaxResolution(req, family).toUpperCase(),
     prompt_expansion_mode: req.promptExpansion ?? "balanced"
   };
   if (req.safetyChecker !== void 0) input.enable_safety_checker = req.safetyChecker;
@@ -5049,11 +5076,14 @@ var WAN_MODEL_IDS = /* @__PURE__ */ new Set(["alibaba/wan-3.0"]);
 function isWan(providerModelId) {
   return WAN_MODEL_IDS.has(providerModelId.toLowerCase());
 }
+function wanResolution(req) {
+  return req.resolution ?? "1080p";
+}
 function wanVideo(req, family) {
   const input = {
     prompt: req.prompt.trim(),
     duration: req.durationSeconds ?? 5,
-    resolution: req.resolution ?? "1080p",
+    resolution: wanResolution(req),
     aspect_ratio: req.aspectRatio ?? "adaptive",
     audio: req.generateAudio ?? true
   };
@@ -5652,7 +5682,7 @@ async function downloadVideo(ctx, url, apiKey, signal) {
     mime: res.headers.get("content-type")?.split(";")[0]?.trim() || "video/mp4"
   };
 }
-function videoUsage2(ctx, req, duration) {
+function videoUsage2(ctx, duration, resolution) {
   const event = {
     model: ctx.model.id,
     provider: "google-genai",
@@ -5661,7 +5691,7 @@ function videoUsage2(ctx, req, duration) {
     quantity: duration,
     durationSeconds: duration
   };
-  if (req.resolution !== void 0) event.resolution = req.resolution;
+  if (resolution !== void 0) event.resolution = resolution;
   return event;
 }
 async function generateVeoVideo(req, ctx, createGenAIClient) {
@@ -5755,7 +5785,7 @@ async function generateVeoVideo(req, ctx, createGenAIClient) {
     }
   };
   const jobId = operation.name ?? crypto.randomUUID();
-  const usage2 = videoUsage2(ctx, req, duration);
+  const usage2 = videoUsage2(ctx, duration, req.resolution ?? "720p");
   return {
     jobId,
     async poll() {
@@ -5879,10 +5909,8 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
   if (req.sourceVideo === void 0 && !extend && req.aspectRatio !== void 0 && req.aspectRatio !== "adaptive") {
     responseFormat.aspect_ratio = toOmniAspect(req.aspectRatio);
   }
-  if (!isOmniV1) {
-    const omniResolution = toOmniResolution(req.resolution);
-    if (omniResolution !== void 0) responseFormat.resolution = omniResolution;
-  }
+  const omniResolution = isOmniV1 ? void 0 : toOmniResolution(req.resolution);
+  if (omniResolution !== void 0) responseFormat.resolution = omniResolution;
   if (req.durationSeconds !== void 0) responseFormat.duration = `${req.durationSeconds}s`;
   const client2 = createGenAIClient({
     apiKey,
@@ -6000,7 +6028,7 @@ async function generateOmniVideo(req, ctx, createGenAIClient) {
     if (!(extend && !chained)) await cleanupUploads();
   }
   const jobId = interaction.id ?? crypto.randomUUID();
-  const usage2 = videoUsage2(ctx, req, duration);
+  const usage2 = videoUsage2(ctx, duration, omniResolution ?? "720p");
   const getInteraction = async () => {
     try {
       const getParams = { stream: false };
@@ -7582,7 +7610,8 @@ function videoUsage4(ctx, req) {
     unit: "compute_seconds",
     quantity: 0
   };
-  if (req.resolution !== void 0) event.resolution = req.resolution;
+  const resolution = replicateVideoResolution(ctx.model.providerModelId, req);
+  if (resolution !== void 0) event.resolution = resolution;
   if (req.durationSeconds !== void 0) event.durationSeconds = req.durationSeconds;
   return event;
 }
@@ -7819,7 +7848,7 @@ function klingInput(req) {
   const input = {
     prompt: req.prompt,
     duration: req.durationSeconds ?? 10,
-    mode: req.mode ?? "standard",
+    mode: klingTier(req).mode,
     generate_audio: req.generateAudio ?? false
   };
   const startImage = images[0];
@@ -7833,17 +7862,25 @@ function klingInput(req) {
   if (req.negativePrompt !== void 0) input.negative_prompt = req.negativePrompt;
   return input;
 }
-function klingOmniMode(req) {
-  if (req.resolution === "4k") return "4k";
-  if (req.resolution === "1080p") return "pro";
-  if (req.resolution === "720p") return "standard";
-  return req.mode ?? "standard";
+function replicateVideoResolution(providerModelId, req) {
+  const kind = videoKind(providerModelId);
+  if (kind === "kling" || kind === "kling-omni") return klingTier(req).resolution;
+  if (kind === "hailuo") return hailuoResolution(req);
+  return req.resolution;
+}
+var KLING_TIERS = [
+  { mode: "standard", resolution: "720p" },
+  { mode: "pro", resolution: "1080p" },
+  { mode: "4k", resolution: "4k" }
+];
+function klingTier(req) {
+  return KLING_TIERS.find((tier) => tier.resolution === req.resolution) ?? KLING_TIERS.find((tier) => tier.mode === req.mode) ?? KLING_TIERS[0];
 }
 function klingOmniInput(req) {
   const input = {
     prompt: req.prompt,
     duration: req.durationSeconds ?? 10,
-    mode: klingOmniMode(req),
+    mode: klingTier(req).mode,
     generate_audio: req.generateAudio ?? false
   };
   const start = req.firstFrame ?? req.lastFrame;
@@ -7857,11 +7894,14 @@ function klingOmniInput(req) {
   if (refs.length > 0) input.reference_images = refs;
   return input;
 }
+function hailuoResolution(req) {
+  return req.resolution ?? "768p";
+}
 function hailuoInput(req) {
   const input = {
     prompt: req.prompt,
     duration: req.durationSeconds ?? 6,
-    resolution: req.resolution ?? "768p",
+    resolution: hailuoResolution(req),
     prompt_optimizer: true
   };
   if (req.firstFrame) input.first_frame_image = refToUrl(req.firstFrame);
@@ -9689,6 +9729,33 @@ function createModelKit(config, adapters) {
       result: () => done
     };
   }
+  function wrapVideoJob(model, job, warnings, startedAt) {
+    const result = job.result.bind(job);
+    let settled;
+    return {
+      ...job,
+      warnings,
+      result(opts) {
+        if (settled) return settled;
+        const pending = result(opts).then(
+          async (out) => {
+            const usage2 = timed(
+              { ...out.usage, model: model.id, unit: model.usage.unit },
+              startedAt
+            );
+            await emit(usage2);
+            return { ...out, usage: usage2 };
+          },
+          (cause) => {
+            settled = void 0;
+            throw cause;
+          }
+        );
+        settled = pending;
+        return pending;
+      }
+    };
+  }
   return {
     catalog,
     describe: (id) => catalog.require(id),
@@ -9773,31 +9840,52 @@ function createModelKit(config, adapters) {
         });
       const startedAt = Date.now();
       const job = await adapter.generateVideo(request, ctx);
-      const result = job.result.bind(job);
-      let settled;
-      return {
-        ...job,
-        warnings,
-        result(opts) {
-          if (settled) return settled;
-          const pending = result(opts).then(
-            async (out) => {
-              const usage2 = timed(
-                { ...out.usage, model: model.id, unit: model.usage.unit },
-                startedAt
-              );
-              await emit(usage2);
-              return { ...out, usage: usage2 };
-            },
-            (cause) => {
-              settled = void 0;
-              throw cause;
-            }
-          );
-          settled = pending;
-          return pending;
-        }
-      };
+      return wrapVideoJob(model, job, warnings, startedAt);
+    },
+    async resumeVideo(req, jobId, call = {}) {
+      mediaRetryOptions(call);
+      if (!jobId) {
+        throw new ModelKitError({
+          code: "missing_input",
+          message: "resumeVideo needs a jobId",
+          retryable: false
+        });
+      }
+      if (!isCatalogId(req.model)) {
+        throw new ModelKitError({
+          code: "unsupported_option",
+          message: "resumeVideo needs a catalog model id"
+        });
+      }
+      const model = catalog.require(req.model);
+      if (model.type !== "video")
+        throw new ModelKitError({
+          code: "unsupported_option",
+          message: `${model.id} is not a video model`,
+          model: model.id
+        });
+      const { request: input, warnings: inputWarnings } = videoRequestOf({
+        ...req,
+        model: req.model
+      });
+      const { request, warnings } = resolveVideoOptions(
+        model,
+        input,
+        call.onUnsupported ?? "error"
+      );
+      warnings.push(...inputWarnings);
+      logCoercions(logger, model.id, warnings);
+      const signal = combineSignals(req.abortSignal, call.signal);
+      const { adapter, ctx } = context(model, signal ? { ...call, signal } : call, warnings);
+      if (!adapter.resumeVideo)
+        throw new ModelKitError({
+          code: "unsupported_option",
+          message: `${adapter.id} cannot resume a video job`,
+          model: model.id
+        });
+      const startedAt = Date.now();
+      const job = adapter.resumeVideo(request, jobId, ctx);
+      return wrapVideoJob(model, { ...job, jobId }, warnings, startedAt);
     },
     async generateAudio(req, call = {}) {
       mediaRetryOptions(call);
@@ -10042,6 +10130,7 @@ var describe = (modelId) => createDefaultModelKit().describe(modelId);
   SurfaceSchema,
   VIDEO_INPUT_SPEC_KEYS,
   VIDEO_TOOL_MODELS,
+  VideoResolutionSchema,
   adapterServesType,
   aiSdkTextAdapter,
   arkAdapter,