# TopK

> TopK is an AI-native search engine and document database with native multi-vector, keyword, and faceted search combined in a single composable query.

TopK is an AI-native search engine built for scale and performance, combining dense (vector) and sparse (keyword) retrieval in a single composable query. It serves as a document database with native multi-vector, keyword, and faceted search, designed for modern AI applications in production. TopK is built for enterprise workloads, offering SOC 2 Type I compliance, 99.9% SLA, and deployment options in your own VPC or on the public cloud.

- **True Hybrid Retrieval™** — *Combines dense (vector) and sparse (keyword) retrieval to deliver accurate top-k results; get started by querying with `fn.semantic_similarity` and `fn.bm25_score` in a single call.*
- **Fast, High-Recall Filtering** — *Innovative filtering outpaces traditional vector databases in both query speed and result completeness; apply filter expressions directly in your query.*
- **Flexible Scoring** — *A powerful expression language lets you weight and combine multiple signals (semantic, keyword, vector distance) based on business logic or domain bias.*
- **Multi-Modal Search** — *Supports unlimited vector representations per document, enabling semantic search across text, images, video, and audio.*
- **Large-Scale & Multi-Tenant** — *Scales to millions (and billions) of documents in single and multi-tenant scenarios using simple filter expressions.*
- **Easy Integration via SDKs** — *Official SDKs for Python, JavaScript, and Rust; connectors for Postgres, MongoDB, and Kafka; get started by installing the SDK and pointing it at your TopK collection.*
- **Bring Your Own Cloud** — *Deploy on AWS, GCP, or Azure in your own VPC, or use TopK's managed public cloud offering.*
- **Provisioned Capacity Pricing** — *Usage-based pricing model covering storage, writes, and queries ensures predictable and scalable costs.*
- **Enterprise Security** — *SOC 2 Type I compliant with 99.9% availability SLA and 24/7 on-call support via a private Slack channel.*
- **Sub-10ms Latency at Scale** — *Benchmarks show consistent p50 latency of ~3.91ms and >98% recall at 1M+ documents.*

## Features
- True Hybrid Retrieval (dense + sparse)
- Multi-vector search per document
- BM25 keyword search
- Faceted and metadata filtering
- Flexible expression-based scoring
- Multi-tenant support
- Sub-10ms query latency
- High-recall filtering
- Multi-modal search (text, image, video, audio)
- Bring Your Own Cloud (AWS, GCP, Azure)
- Provisioned capacity pricing
- SOC 2 Type I compliance
- 99.9% availability SLA
- 24/7 on-call support
- Private Slack channel support

## Integrations
Postgres, MongoDB, Kafka, Python, JavaScript, Rust, AWS, GCP, Azure

## Platforms
WEB, API, DEVELOPER_SDK, CLI

## Pricing
Paid

## Links
- Website: https://www.topk.io
- Documentation: https://docs.topk.io/
- Repository: https://github.com/topk-io/topk
- EveryDev.ai: https://www.everydev.ai/tools/topk