DSPy-Qdrant

Qdrant powered custom retriever module for DSPy.

Installation

pip install dspy-qdrant

Usage

from dspy_qdrant import QdrantRM

The QdrantRM class enables semantic search by retrieving the top-k most relevant documents from a Qdrant collection using a sentence embedding model.

Parameters

Name	Type	Description	Default
qdrant_collection_name	str	Name of the Qdrant collection used for retrieval.	Required
qdrant_client	QdrantClient	An initialized instance of qdrant_client.QdrantClient.	Required
k	int	Number of top documents to retrieve per query.	3
document_field	str	Field in the Qdrant payload that contains the raw document content.	"document"
vectorizer	BaseSentenceVectorizer	Embedding model for vectorizing queries. Uses FastEmbedVectorizer if not provided.	None
vector_name	str	Name of the vector field in Qdrant collection to use for search. Defaults to the first found.	None

Use in a Module's forward() Function

import dspy

from qdrant_client import QdrantClient
from dspy_qdrant import QdrantRM

qdrant_client = QdrantClient()

class MyModule(dspy.Module):
    def __init__(self, num_passages: int = 5):
        super().__init__()
        self.num_passages = num_passages

    def forward(self, question: str):
        retrieve = QdrantRM(
            qdrant_collection_name="my_collection_name",
            qdrant_client=qdrant_client,
            k=self.num_passages
        )
        # Do something with results...

📚 See Also

• Qdrant Documentation
• DSPy GitHub