climage 0.5.0 → 0.5.1

package/dist/cli.js

@@ -653,6 +653,9 @@ var falProvider = {
 
 // src/providers/google.ts
 import { GoogleGenAI } from "@google/genai";
+import { mkdtemp, readFile, rm } from "fs/promises";
+import { tmpdir } from "os";
+import { join } from "path";
 function getGeminiApiKey(env) {
   return env.GEMINI_API_KEY || env.GOOGLE_API_KEY || env.GOOGLE_GENAI_API_KEY;
 }
@@ -700,6 +703,21 @@ function imageToGoogleFormat(imageInput) {
   }
   return { fileUri: imageInput };
 }
+function imageToVeoFormat(imageInput) {
+  if (imageInput.startsWith("data:")) {
+    const parsed = parseDataUri(imageInput);
+    if (!parsed?.data) {
+      throw new Error("Failed to parse data URI for Veo image input");
+    }
+    return { imageBytes: parsed.data, mimeType: parsed.mimeType };
+  }
+  if (imageInput.startsWith("gs://")) {
+    return { gcsUri: imageInput };
+  }
+  throw new Error(
+    `Veo image inputs must be data: URIs or gs:// URIs (got ${imageInput.slice(0, 24)}...)`
+  );
+}
 var GEMINI_IMAGE_MODELS = ["gemini-2.5-flash-image", "gemini-3-pro-image-preview"];
 function resolveModel(model) {
   if (!model) return "gemini-2.5-flash-image";
@@ -718,6 +736,31 @@ async function downloadBytes3(url) {
   log3(`Downloaded ${ab.byteLength} bytes in ${Date.now() - start}ms, type: ${ct}`);
   return { bytes: new Uint8Array(ab), mimeType: ct };
 }
+async function downloadGeneratedVideo(ai, generatedVideo) {
+  const video = generatedVideo?.video;
+  if (video?.videoBytes) {
+    return {
+      bytes: new Uint8Array(Buffer.from(video.videoBytes, "base64")),
+      mimeType: video.mimeType
+    };
+  }
+  if (video?.uri && !video.uri.startsWith("gs://")) {
+    try {
+      return await downloadBytes3(video.uri);
+    } catch (err) {
+      log3("Direct video download failed, falling back to ai.files.download:", String(err));
+    }
+  }
+  const tempDir = await mkdtemp(join(tmpdir(), "climage-veo-"));
+  const downloadPath = join(tempDir, "video.mp4");
+  try {
+    await ai.files.download({ file: generatedVideo, downloadPath });
+    const buf = await readFile(downloadPath);
+    return { bytes: new Uint8Array(buf), mimeType: video?.mimeType ?? "video/mp4" };
+  } finally {
+    await rm(tempDir, { recursive: true, force: true });
+  }
+}
 async function sleep2(ms) {
   await new Promise((r) => setTimeout(r, ms));
 }
@@ -783,15 +826,15 @@ async function generateWithVeo(ai, model, req) {
   const config = {
     numberOfVideos: req.n,
     ...req.aspectRatio ? { aspectRatio: req.aspectRatio } : {},
-    // Add duration if specified (Veo 3.1 supports 4, 6, 8)
-    ...req.duration !== void 0 ? { durationSeconds: String(req.duration) } : {}
+    // Add duration if specified (Veo supports 4-8 seconds depending on model)
+    ...req.duration !== void 0 ? { durationSeconds: req.duration } : {}
   };
   if (req.inputImages?.length && isVeo31Model(model)) {
     const referenceImages = req.inputImages.slice(0, 3).map((img) => {
-      const imageData = imageToGoogleFormat(img);
+      const imageData = imageToVeoFormat(img);
       return {
         image: imageData,
-        referenceType: "asset"
+        referenceType: "ASSET"
       };
     });
     config.referenceImages = referenceImages;
@@ -804,12 +847,12 @@ async function generateWithVeo(ai, model, req) {
   };
   const firstFrameImage = req.startFrame ?? (req.inputImages?.length === 1 ? req.inputImages[0] : void 0);
   if (firstFrameImage && isVeo31Model(model)) {
-    const imageData = imageToGoogleFormat(firstFrameImage);
+    const imageData = imageToVeoFormat(firstFrameImage);
     generateParams.image = imageData;
     log3("Added first frame image");
   }
   if (req.endFrame && isVeo31Model(model)) {
-    const lastFrameData = imageToGoogleFormat(req.endFrame);
+    const lastFrameData = imageToVeoFormat(req.endFrame);
     config.lastFrame = lastFrameData;
     log3("Added last frame for interpolation");
   }
@@ -839,26 +882,22 @@ async function generateWithVeo(ai, model, req) {
   for (let i = 0; i < Math.min(videos.length, req.n); i++) {
     const v = videos[i];
     log3(`Processing video ${i}:`, JSON.stringify(v).slice(0, 300));
-    const uri = v?.video?.uri;
-    if (!uri) {
-      log3(`Video ${i} has no URI, skipping`);
+    if (!v?.video) {
+      log3(`Video ${i} has no video payload, skipping`);
       continue;
     }
-    if (uri.startsWith("gs://")) {
-      throw new Error(
-        `Google Veo returned a gs:// URI (${uri}). Configure outputGcsUri / Vertex flow to fetch from GCS.`
-      );
-    }
-    const { bytes, mimeType } = await downloadBytes3(uri);
-    out.push({
+    const uri = v?.video?.uri;
+    const { bytes, mimeType } = await downloadGeneratedVideo(ai, v);
+    const item = {
       kind: "video",
       provider: "google",
       model,
       index: i,
-      url: uri,
       bytes,
       ...mimeType !== void 0 ? { mimeType } : {}
-    });
+    };
+    if (uri) item.url = uri;
+    out.push(item);
   }
   if (!out.length) throw new Error("Google Veo returned videos but none were downloadable");
   log3(`Successfully generated ${out.length} video(s)`);