Skip to content

Qdrant

Qdrant is a high-performance vector database (self-hosted or Qdrant Cloud). It requires an embedder because it stores and searches dense vectors you supply.

When should you use this?

Use it when you already run Qdrant in production or need a scalable vector store.

Prerequisites

  • Install the client: 
    pip install "openagent-eval[qdrant]"
  • An embedder (e.g. sentence_transformers) — required.
  • A Qdrant instance (local :memory:, Docker, or Cloud).

Step 1 — Install

pip install "openagent-eval[qdrant]"

Step 2 — Configure

config.yaml
retriever:
  provider: qdrant
  settings:
    collection_name: my_docs
    url: http://localhost:6333
    # api_key: <your-qdrant-key>     # for Qdrant Cloud
    distance: Cosine                  # Cosine | Euclid | Dot
  embedder:
    provider: sentence_transformers
    model: all-MiniLM-L6-v2

Step 3 — Run

oaeval run config.yaml

Python SDK example

eval_qdrant.py
import asyncio

from openagent_eval.config.models import Config, RetrieverConfig, EmbedderConfig
from openagent_eval.core.engine import Engine

config = Config(
    dataset={"path": "data/questions.json"},
    llm={"provider": "mock"},
    retriever=RetrieverConfig(
        provider="qdrant",
        settings={
            "collection_name": "my_docs",
            "url": "http://localhost:6333",
            "distance": "Cosine",
        },
        embedder=EmbedderConfig(
            provider="sentence_transformers", model="all-MiniLM-L6-v2"
        ),
    ),
    metrics={"retrieval": ["context_precision", "context_recall", "mrr"]},
)

engine = Engine(config)
report = asyncio.run(engine.run(dataset))
print(report.summary["metrics_summary"])

All configuration options

Option Type Default Description
collection_name str — Required. Qdrant collection.
embedder Embedder — Required. Injected automatically from retriever.embedder.
url str \| null null Qdrant URL; null → in-memory :memory:.
api_key str \| null null For Qdrant Cloud.
prefer_grpc bool False Use gRPC transport.
distance str Cosine Cosine, Euclid, or Dot.

Troubleshooting

  • ProviderConnectionError: embedder is required — add retriever.embedder.
  • Connection failed — verify url and api_key (Cloud).