Video Semantic Search Pipeline

from mixpeek import Mixpeek

client = Mixpeek(api_key="YOUR_API_KEY")

# 1. Create namespace & collection
namespace = client.namespaces.create(name="video-search")
collection = client.collections.create(
    namespace_id=namespace.id,
    name="videos",
    extractors=["video-embedding-v2", "transcript-extraction"],
    chunk_strategy="scene-based"
)

# 2. Upload videos
client.buckets.upload(
    collection_id=collection.id,
    url="s3://your-bucket/videos/"
)

# 3. Create retriever
retriever = client.retrievers.create(
    namespace_id=namespace.id,
    name="video-search",
    stages=[
        {"type": "vector_search", "model": "multilingual-e5-large", "top_k": 50},
        {"type": "rerank", "model": "colbert-v2", "top_k": 10}
    ]
)

# 4. Search
results = client.retrievers.execute(
    retriever_id=retriever.id,
    query="person explaining machine learning concepts"
)