> ## Documentation Index
> Fetch the complete documentation index at: https://infino.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# LlamaIndex integration with Infino vector store

> Use Infino as a LlamaIndex vector store for vector, BM25 full-text, hybrid (RRF), and SQL retrieval over one copy of your data on object storage.

Infino is a LlamaIndex vector store: vector, full-text (BM25), hybrid (RRF), and
SQL retrieval over a single copy of your data on object storage. It stores the
node text alongside the embedding, so one Infino table replaces both the vector
store and the docstore. Bring your own embeddings.

```bash theme={null}
pip install llama-index-vector-stores-infino
```

The [`llama-index-vector-stores-infino`](https://pypi.org/project/llama-index-vector-stores-infino/)
package provides `InfinoVectorStore`, a `BasePydanticVectorStore` with
`stores_text=True`. Requires Python 3.10+; bring an embeddings provider
separately (for example `pip install llama-index-embeddings-openai`).

## Quickstart

```python theme={null}
import infino
from llama_index.core import Document, StorageContext, VectorStoreIndex
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.vector_stores.infino import InfinoVectorStore

connection = infino.connect("./data")
store = InfinoVectorStore(connection, table_name="docs", dim=1536)

storage_context = StorageContext.from_defaults(vector_store=store)
index = VectorStoreIndex.from_documents(
    [Document(text="Infino runs search on object storage.")],
    storage_context=storage_context,
    embed_model=OpenAIEmbedding(),
)

retriever = index.as_retriever(similarity_top_k=4)
```

## Local, object storage, or hosted cloud

The store construction is identical everywhere; only the `infino.connect(...)`
call differs.

```python theme={null}
import infino

# Local (embedded): Parquet under a directory, no server.
connection = infino.connect("./data")

# Object storage (embedded): Parquet in your bucket, queried in place.
connection = infino.connect(
    "s3://my-bucket/rag", storage_options={"aws_region": "us-east-1"}
)

# Hosted Infino Cloud: sign up at https://platform.infino.ws for an API key.
connection = infino.connect(
    "https://api.platform.infino.ws/<database>", api_key="..."
)
```

## Query modes

All four `VectorStoreQueryMode`s are first-class on the same store:

```python theme={null}
index.as_retriever(similarity_top_k=10)                                            # vector (default)
index.as_retriever(similarity_top_k=10, vector_store_query_mode="text_search")     # BM25
index.as_retriever(similarity_top_k=10, vector_store_query_mode="hybrid")          # hybrid (RRF)
index.as_retriever(similarity_top_k=10, vector_store_query_mode="mmr",
                   vector_store_kwargs={"embed_model": embed_model})               # MMR
```

Hybrid fuses BM25 and vector search with reciprocal-rank fusion in one engine
call. Structured metadata filters (the full LlamaIndex operator set), a
text-pushdown pre-filter, a SQL escape hatch, and async variants are all
supported; see the
[package README](https://github.com/infino-ai/llama-index-vector-stores-infino)
for the complete reference.

## See also

* [Quickstart](/docs/quickstart)
* [Search: full-text, vector, hybrid](/docs/guides/search)
* [MCP server](/docs/integrations/mcp)