Dataset Versioning

from mixpeek import Mixpeek

client = Mixpeek(api_key="YOUR_API_KEY", namespace="training-data")

# 1. Every object records the dataset version it belongs to
bucket = client.buckets.create(
    bucket_name="training-data",
    bucket_schema={"properties": {"image": {"type": "image"}}},
)
client.buckets.upload(
    bucket["bucket_id"],
    blobs=[{"property": "image", "type": "image", "data": "s3://your-bucket/training/v2/img-0001.jpg"}],
    metadata={"version": "v2"},
)
client.buckets.upload(
    bucket["bucket_id"],
    blobs=[{"property": "image", "type": "image", "data": "s3://your-bucket/training/v3/img-0001.jpg"}],
    metadata={"version": "v3"},
)

# 2. One collection per version, scoped with a source filter on that metadata
v2 = client.collections.create(
    collection_name="training-v2",
    source={"type": "bucket", "bucket_ids": [bucket["bucket_id"]], "source_filters": {"filters": {"AND": [{"field": "metadata.version", "operator": "eq", "value": "v2"}]}}},
    feature_extractor={
        "feature_extractor_name": "multimodal_extractor",
        "version": "v1",
    },
)

# The next version is a clone with a different filter; extractor and settings carry over
v3 = client.collections.clone(
    v2["collection_id"],
    collection_name="training-v3",
    source={
        "type": "bucket",
        "bucket_ids": [bucket["bucket_id"]],
        "source_filters": {"filters": {"AND": [{"field": "metadata.version", "operator": "eq", "value": "v3"}]}},
    },
)
client.collections.trigger(v2["collection_id"])
client.collections.trigger(v3["collection_id"])

# 3. Query each version on its own and compare
for version in ("training-v2", "training-v3"):
    retriever = client.retrievers.create(
        retriever_name="search-" + version,
        collection_identifiers=[version],
        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 demos"})
    print(version, [doc["document_id"] for doc in results["documents"][:5]])