Face Search Pipeline

from mixpeek import Mixpeek

client = Mixpeek(api_key="YOUR_API_KEY", namespace="face-search")

# 1. A bucket for the video content, and a collection that detects, aligns and embeds every face with ArcFace
bucket = client.buckets.create(
    bucket_name="video-content",
    bucket_schema={
        "properties": {
            "video": {
                "type": "video",
            },
        },
    },
)
collection = client.collections.create(
    collection_name="video_faces",
    source={"type": "bucket", "bucket_ids": [bucket["bucket_id"]]},
    feature_extractor={
        "feature_extractor_name": "face_identity_extractor",
        "version": "v1",
    },
)

# 2. Upload and process
client.buckets.upload(
    bucket["bucket_id"],
    blobs=[{"property": "video", "type": "video", "data": "s3://your-bucket/videos/panel-2026-06.mp4"}],
)
client.collections.trigger(collection["collection_id"])

# 3. An example face searched against the ArcFace index, then labeled from tax_your_known_faces, a flat taxonomy over a collection of reference photos
retriever = client.retrievers.create(
    retriever_name="face-search",
    collection_identifiers=["video_faces"],
    input_schema={
        "face": {
            "type": "image",
            "required": True,
        },
    },
    stages=[
        {
            "stage_name": "search",
            "stage_id": "feature_search",
            "parameters": {
                "searches": [
                    {
                        "feature_uri": "mixpeek://face_identity_extractor@v1/insightface__arcface",
                        "query": {
                            "input_mode": "content",
                            "value": "{{INPUT.face}}",
                        },
                        "top_k": 50,
                    },
                ],
                "final_top_k": 50,
            },
        },
        {
            "stage_name": "label",
            "stage_id": "taxonomy_enrich",
            "parameters": {
                "taxonomy_id": "tax_your_known_faces",
                "top_k": 1,
            },
        },
    ],
)

# 4. Search
results = client.retrievers.execute(
    retriever["retriever_id"],
    inputs={
    "face": "https://example.com/person-photo.jpg",
},
)

for doc in results["documents"]:
    print(doc["timestamp"], doc["score"])