Astra DB integrations

Integrate with Astra DB using LangChain Python.

DataStax Astra DB is a serverless AI-ready database built on Apache Cassandra® and made conveniently available through an easy-to-use JSON API.

See a tutorial provided by DataStax.

Installation and setup

Install the following Python package:

pip

pip install "langchain-astradb>=0.6,<0.7"

uv add langchain-astradb>=0.6,<0.7

Create a database (if needed) and get the connection secrets. Set the following variables:

ASTRA_DB_API_ENDPOINT="API_ENDPOINT"
ASTRA_DB_APPLICATION_TOKEN="TOKEN"

Vector store

A few typical initialization patterns are shown here:

from langchain_astradb import AstraDBVectorStore

vector_store = AstraDBVectorStore(
    embedding=my_embedding,
    collection_name="my_store",
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
)


from astrapy.info import VectorServiceOptions

vector_store_vectorize = AstraDBVectorStore(
    collection_name="my_vectorize_store",
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
    collection_vector_service_options=VectorServiceOptions(
        provider="nvidia",
        model_name="NV-Embed-QA",
    ),
)


from astrapy.info import (
    CollectionLexicalOptions,
    CollectionRerankOptions,
    RerankServiceOptions,
    VectorServiceOptions,
)

vector_store_hybrid = AstraDBVectorStore(
    collection_name="my_hybrid_store",
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
    collection_vector_service_options=VectorServiceOptions(
        provider="nvidia",
        model_name="NV-Embed-QA",
    ),
    collection_lexical=CollectionLexicalOptions(analyzer="standard"),
    collection_rerank=CollectionRerankOptions(
        service=RerankServiceOptions(
            provider="nvidia",
            model_name="nvidia/llama-3.2-nv-rerankqa-1b-v2",
        ),
    ),
)

Notable features of class AstraDBVectorStore:

• native async API;
• metadata filtering in search;
• MMR (maximum marginal relevance) search;
• server-side embedding computation (“vectorize” in Astra DB parlance);
• auto-detect its settings from an existing, pre-populated Astra DB collection;
• hybrid search (vector + BM25 and then a rerank step);
• support for non-Astra Data API (e.g. self-hosted HCD deployments);

Learn more in the example notebook.

See the example provided by DataStax.

LLM cache

from langchain.globals import set_llm_cache
from langchain_astradb import AstraDBCache

set_llm_cache(AstraDBCache(
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
))

Semantic LLM cache

from langchain.globals import set_llm_cache
from langchain_astradb import AstraDBSemanticCache

set_llm_cache(AstraDBSemanticCache(
    embedding=my_embedding,
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
))

Document loader

from langchain_astradb import AstraDBLoader

loader = AstraDBLoader(
    collection_name="my_collection",
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
)

Learn more in the example notebook.

Self-querying retriever

from langchain_astradb import AstraDBVectorStore
from langchain_classic.retrievers.self_query.base import SelfQueryRetriever

vector_store = AstraDBVectorStore(
    embedding=my_embedding,
    collection_name="my_store",
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
)

retriever = SelfQueryRetriever.from_llm(
    my_llm,
    vector_store,
    document_content_description,
    metadata_field_info
)

Store

from langchain_astradb import AstraDBStore

store = AstraDBStore(
    collection_name="my_kv_store",
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
)

See the API Reference for the AstraDBStore.

Byte store

from langchain_astradb import AstraDBByteStore

store = AstraDBByteStore(
    collection_name="my_kv_store",
    api_endpoint=ASTRA_DB_API_ENDPOINT,
    token=ASTRA_DB_APPLICATION_TOKEN,
)

See the API reference for the AstraDBByteStore.

---

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.

Edit this page on GitHub or file an issue.