BlogRead Why Vector DBs Are the Wrong Abstraction – And What We Built Instead

Search engine for the AI era

Built for scale and performance, TopK is a document database with native multi-vector, keyword, and faceted search—all in a single, composable query. It's the starting point for a new class of AI-native search engines—ones that understand your data, eliminate complexity, and raise the bar for relevance in production.

Empower your search with
better
performance

collection("records").query(
select(
"title",
"content",
# semantic (vector) search on "title" field
title_similarity=fn.semantic_similarity("title", "abdominal pain"),
# semantic (vector) search on "content" field
content_similarity=fn.semantic_similarity("content", "abdominal pain"),
# vector search on "image_embedding" field
image_score=fn.vector_distance("image_embedding", [...]),
# collect BM25 score on all text-indexed fields
keyword_score=fn.bm25_score(),
)
.filter(
match("abdominal pain") # keyword search
)
.filter(
field("published_at") >= 2024 && field("status") == "ACTIVE"
)
.topk(
field("title_simlarity") * 0.3 + # 30% weight for title semantic similarity
field("content_similarity") * 0.2 + # 20% weight for content semantic similarity
field("image_score") * 0.1 + # 10% weight for image embedding vector score
field("keyword_score") * 0.4, # 40% weight for keyword score
# get top 10 results
10
)
)

On top of your
existing infrastructure

No need to change your primary database. TopK provides you with tools, support and guidance to integrate with your stack seamlessly, whether deployed in your VPC or in our public cloud.

Integrate TopK with your stack, worry-free.

Connectors

  • Postgres
  • MongoDB
  • Kafka

Support

TopK Diagram

Benchmarks

TopK delivers blazing fast vector search that scales - handling millions of documents with sub-10ms latency at high QPS. Our benchmarks show consistent performance even as collection size grows.

Latency Scaling

(dim=768, k=10, selectivity=100%)
p5024.00 msp9534.00 msp9941.00 ms051015202530354045Latency [ms]

QPS vs Concurrent clients

(dim=768, k=10, selectivity=100%)
05101520253035404550Queries per second1248Concurrent clients

>98%

Recall

For more details about the TopK's performance, please refer to the benchmarks page.

TopK is built for enterprise.

Scalable, Customizable, and Secure — Enterprise‐Grade Compliance, Reliability, and 99.9% SLA.

Multi-Modal Search

Support for unlimited vector representations per document, enabling rich semantic search across text, images, video and audio.

Bring Your Own Cloud

Deploy on your preferred cloud provider (AWS, GCP, Azure) or use our public cloud offering.

Provisioned Capacity

Provisioned capacity pricing model ensures predictable costs and scalable performance.

High Write Throughput

Sustained high-performance write operations for demanding workloads.

Easy Integration

SDKs for Python, JavaScript, and Rust, plus connectors for Postgres, MongoDB, and Kafka.

Enterprise Security

SOC 2 Type I compliant with 99.9% availability SLA.

Ready to get started?

TopK provides a powerful and flexible search engine built for modern AI applications, combining vector search, keyword search and metadata filtering in a single query.