---
id: upstash-vector
version: "0.8.0"
license: MIT
license_treatment: permissive
maintenance: aging
---

# upstash-vector — Serverless Vector SDK from Upstash

License: permissive · Maintenance: aging · Downloads: 354.3K/mo

## What it is and what it does

Upstash Vector is a Python client for Upstash's serverless vector database service. It provides a REST-based interface to store and query vectors without managing your own infrastructure. The package handles three index types—dense (traditional embeddings), sparse (keyword-based), and hybrid (combining both)—and supports optional metadata filtering and associated data fields.

You initialize an Index with credentials from the Upstash console, then call upsert() to insert or update vectors and query() to find similar vectors by similarity score. The client abstracts away HTTP details via httpx and supports multiple input formats (tuples, dicts, or Vector objects). It is classified as Alpha in development status but marked GA by Upstash and backed by their professional support.

Use it for:

- Store and retrieve embeddings from language models for semantic search applications.
- Build recommendation systems by querying similar vectors based on user or item embeddings.
- Implement metadata-filtered vector search to narrow results by categorical or structured fields.
- Use with embedding models to automatically vectorize and store raw text data without manual embedding.
- Prototype vector-based features without provisioning and managing a dedicated vector database.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Python client for Upstash Vector, a serverless vector database—upserts and queries vectors with optional metadata and data fields, supporting dense, sparse, and hybrid index types.

Yes, if you are already using Upstash Vector or committed to a serverless vector database architecture. The package is straightforward, has low install friction, and is backed by professional support. However, the aging maintenance status (533 days since last release) and Alpha classification warrant caution for mission-critical production systems; verify that your use case aligns with Upstash's roadmap and support commitments.

## Install

pip install upstash-vector
uv add upstash-vector
poetry add upstash-vector

## Installing upstash-vector

Before you install: Low install friction with a single runtime dependency (httpx). Maintenance status is aging—last commit was 2025-10-21 and the package has not been updated in 533 days, though the project is marked GA and receives Upstash Professional Support.

License in practice: MIT license is permissive, allowing commercial and private use with minimal restrictions.

Quickstart:

pip install upstash-vector

from upstash_vector import Index

index = Index.from_env()  # Reads UPSTASH_VECTOR_REST_URL and UPSTASH_VECTOR_REST_TOKEN
index.upsert(vectors=[("id1", [0.1, 0.2])])
res = index.query(vector=[0.6, 0.9], top_k=5)

Requires UPSTASH_VECTOR_REST_URL and UPSTASH_VECTOR_REST_TOKEN environment variables or explicit URL and token parameters; an active Upstash Vector database must exist.

Verify before relying:

- Whether the aging maintenance status (533 days since last release) affects stability or feature completeness for production use.
- Performance characteristics and throughput limits for large-scale vector operations.
- Whether sparse and hybrid index types are fully tested and production-ready.

## Package facts

- License: MIT (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: aging
- Downloads: 354.3K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

vector database client, serverless vector storage, semantic search sdk, embedding storage and retrieval, vector similarity search, upstash vector python, managed vector db, vector-search, serverless, embeddings

[View on SkillFed](https://skillfed.io/packages/upstash-vector) · [View on PyPI](https://pypi.org/project/upstash-vector/)