Embeddings + database + ranking in a single API.

TopK combines embeddings with reranking and a powerful query engine in a single API to drastically simplify your search stack.

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

Unified search

Storing documents

Managing collections

Semantic search with keyword-based filtering and reranking in a single API call. No need to integrate multiple providers or build data pipelines.

from topk.query import fn, select, match

docs = client.collection("books").query(
    select(
        "title",
        title_similarity=fn.semantic_similarity("title", "catcher in the rye")
    )
    .filter(match("salinger"))
    .top_k(field("title_similarity"), 10)
    .rerank()
)

Unified search

Storing documents

Managing collections

Semantic search with keyword-based filtering and reranking in a single API call. No need to integrate multiple providers or build data pipelines.

from topk_sdk.query import select, filter, field, fn, match

results = client.collection("books").query(
    select(
        "title",
        vector_distance=fn.vector_distance("title_embedding"),
        text_score=fn.keyword_score(),
    )
        # Keyword search - filter books with word "catcher" in the title
        .filter(field("title").match("catcher"))
        # Vector search - compute vector distance
        .filter(field("vector_distance") > 0.5),
        # Metadata search - filter books by published year
        .filter(field("published_year") > 1920)
        # Scoring - return top 10 results by calculating score as `text_score + vector_distance * 3`
        .top_k(field("text_score") + field("vector_distance") * 3, 10)

    # Pass the LSN to make sure the query is consistent with the writes. 
    lsn=lsn,
)

One-stop shop for all your search needs.

Developer experience is one of the three things we care about at core. Implement TopK with zero effort and focus on what’s the most important - building great products.

Quickstart (5m)

Explore docs

Vector search on dense or sparse vectors with native scalar/binary quantization support.

Index text documents and find the most relevant ones using BM25 scoring.

Hybrid multi-vector + text retrieval to get the most relevant results for your users.

Efficient and flexible filtering to only get the results that matter.

Works seamlessly with your tech stack

Unbeatable affordability, even under heavy loads.

Building from first principles allowed us to be incredibly competitive in terms of price. This allows you to get an edge over your competitors and secure an unbeatable competitive advantage.

Scale to billions of documents without breaking a sweat.

Rely on 99.9+% availability in multiple regions around the world.

Or deploy to your own cloud and your data never leaves your servers.

Built from the ground up to keep customer data isolated and secure.

Use case

RAG

Seamlessly integrate our hybrid retrieval capabilities to enhance generative AI models with real-time, accurate, and relevant data.

Use case

Semantic search

Power your search solution with advanced AI query understanding understanding to deliver precise, meaningful results from unstructured datasets.

Use case

Multi-modal search

Leverage TopK’s multi-vector retrieval capabilities to search across diverse data formats like text, images, audio and video in a single query.

Use case

Recommendations

Enhance user engagement recommendation engine, dynamically tailored to your unique data and user behavior.

White Glove Onboarding

TopK is invitation only.

We want to ensure the best possible experience for you so we'll evaluate your use-case together and get your hands on our product as soon as possible.

White Glove Onboarding

TopK is invitation only.

We want to ensure the best possible experience for you so we'll evaluate your use-case together and get your hands on our product as soon as possible.

White Glove Onboarding

TopK is invitation only.

We want to ensure the best possible experience for you so we'll evaluate your use-case together and get your hands on our product as soon as possible.