Video Keyframe Extraction Pipeline

from mixpeek import Mixpeek

client = Mixpeek(api_key="YOUR_API_KEY", namespace="keyframes")

# 1. A bucket for the videos, and a collection that cuts at scene changes and
#    writes a thumbnail and a multimodal embedding for every scene
bucket = client.buckets.create(
    bucket_name="product-videos",
    bucket_schema={"properties": {"video": {"type": "video"}}},
)
collection = client.collections.create(
    collection_name="product-videos",
    source={"type": "bucket", "bucket_ids": [bucket["bucket_id"]]},
    feature_extractor={
        "feature_extractor_name": "multimodal_extractor",
        "version": "v1",
        "parameters": {
            "split_method": "scene",
            "scene_detection_threshold": 0.5,
            "enable_thumbnails": True,
        },
    },
)

# 2. Upload and process
client.buckets.upload(
    bucket["bucket_id"],
    blobs=[{"property": "video", "type": "video", "data": "s3://your-bucket/product-videos/unboxing.mp4"}],
)
client.collections.trigger(collection["collection_id"])

# 3. Search the scenes visually
retriever = client.retrievers.create(
    retriever_name="keyframe-search",
    collection_identifiers=["product-videos"],
    input_schema={"query": {"type": "text", "required": True}},
    stages=[
        {
            "stage_name": "search",
            "stage_id": "feature_search",
            "parameters": {
                "searches": [
                    {
                        "feature_uri": "mixpeek://multimodal_extractor@v1/vertex_multimodal_embedding",
                        "query": {"input_mode": "text", "value": "{{INPUT.query}}"},
                        "top_k": 20,
                    },
                ],
                "final_top_k": 20,
            },
        },
    ],
)

results = client.retrievers.execute(retriever["retriever_id"], inputs={"query": "product being unboxed"})
for doc in results["documents"]:
    print(doc["thumbnail_url"], doc["start_time"], doc["score"])