respan-instrumentation-qdrant

Respan instrumentation for synchronous and asynchronous qdrant-client operations. It emits canonical Respan task spans for collection management, point writes, payload updates, retrieval, filtering, vector queries, facets, snapshots, and cluster administration.

Install

pip install respan-ai respan-instrumentation-qdrant qdrant-client

Quickstart

from qdrant_client import QdrantClient, models
from respan import Respan, workflow
from respan_instrumentation_qdrant import QdrantInstrumentor

respan = Respan(instrumentations=[QdrantInstrumentor()])


@workflow(name="qdrant_local_query_workflow")
def run_query():
    client = QdrantClient(":memory:")
    client.create_collection(
        "docs",
        vectors_config=models.VectorParams(
            size=3,
            distance=models.Distance.COSINE,
        ),
    )
    client.upsert(
        "docs",
        points=[models.PointStruct(id=1, vector=[0.1, 0.2, 0.3])],
    )
    return client.query_points(
        "docs",
        query=[0.1, 0.2, 0.3],
        limit=1,
    )


print(run_query())
respan.shutdown()

QdrantInstrumentor(capture_content=False) keeps operation names, status, and database attributes while omitting request and response bodies. Activation and deactivation are idempotent, and instrument() / uninstrument() aliases are also available.

See respan-example-projects/python/tracing/qdrant for collection, point, query, async, and deterministic error examples.