Upstash Vector Database

Setup

uv pip install -U upstash-vector pypdf openai agno

Create an index in the Upstash Console, then copy its REST URL and token.

export UPSTASH_VECTOR_REST_URL="your-index-url"
export UPSTASH_VECTOR_REST_TOKEN="your-index-token"

The example uses OpenAI for the agent model, so set your API key:

export OPENAI_API_KEY=xxx

Example

agent_with_knowledge.py

import os

from agno.agent import Agent
from agno.knowledge.knowledge import Knowledge
from agno.vectordb.upstashdb import UpstashVectorDb

vector_db = UpstashVectorDb(
    url=os.getenv("UPSTASH_VECTOR_REST_URL"),
    token=os.getenv("UPSTASH_VECTOR_REST_TOKEN"),
)

knowledge = Knowledge(
    name="Upstash Knowledge Base",
    vector_db=vector_db,
)

knowledge.insert(
    name="Recipes",
    url="https://agno-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf",
    metadata={"doc_type": "recipe_book"},
)

agent = Agent(knowledge=knowledge)
agent.print_response("How to make Pad Thai?", markdown=True)

Without an embedder, UpstashVectorDb uses Upstash’s hosted embedding models. Create your index with an embedding model in the Upstash Console. To embed locally instead, pass an embedder and make sure its dimension matches the index.

Async Support ⚡

Upstash also supports asynchronous operations, enabling concurrency and leading to better performance.

async_upstash_db.py

import asyncio
import os

from agno.agent import Agent
from agno.knowledge.embedder.openai import OpenAIEmbedder
from agno.knowledge.knowledge import Knowledge
from agno.vectordb.upstashdb import UpstashVectorDb

vector_db = UpstashVectorDb(
    url=os.getenv("UPSTASH_VECTOR_REST_URL"),
    token=os.getenv("UPSTASH_VECTOR_REST_TOKEN"),
    dimension=1536,
    embedder=OpenAIEmbedder(enable_batch=True),
)

knowledge = Knowledge(
    vector_db=vector_db,
)

agent = Agent(knowledge=knowledge)

if __name__ == "__main__":
    asyncio.run(
        knowledge.ainsert(
            url="https://agno-public.s3.amazonaws.com/recipes/ThaiRecipes.pdf"
        )
    )

    asyncio.run(agent.aprint_response("How to make Tom Kha Gai", markdown=True))

Use ainsert() and aprint_response() methods with asyncio.run() for non-blocking operations in high-throughput applications.

UpstashVectorDb Params

Parameter	Type	Default	Description
`url`	`str`	-	The Upstash Vector database URL.
`token`	`str`	-	The Upstash Vector API token.
`retries`	`int`	`3`	Number of retry attempts for operations.
`retry_interval`	`float`	`1.0`	Time between retries, in seconds.
`dimension`	`int`	`None`	The dimension of the embeddings. Validated against the index dimension on first use.
`embedder`	`Embedder`	`None`	The embedder to use. When `None`, the index must use an Upstash hosted embedding model.
`namespace`	`str`	`""`	The namespace to use.
`reranker`	`Reranker`	`None`	The reranker to use.
`name`	`str`	`None`	Name of the vector database.
`description`	`str`	`None`	Description of the vector database.

Get Started

Basics

Advanced

Production

Providers

Other

Additional Resources

Upstash Vector Database

Setup

Example

Async Support ⚡

UpstashVectorDb Params

​Setup

​Example

Async Support ⚡

​UpstashVectorDb Params

Setup

Example

UpstashVectorDb Params