Elastic was too time-consuming to maintain, and we wanted a more down-to-earth solution with an S3-compatible backend. We were looking to write our own Tantivy implementation, but Quickwit was released, so we decided to give it a shot. Ever since it has been our fastest AND cheapest log management solution. Additionally, the compatibility with vector.dev was just the cherry on top of the cake!

At Rho, we needed to collect our application traces with no sampling, and to retain them for long periods of time. By providing a blazing fast search over cheap object storage, Quickwit has saved us up to 90% of our tracing telemetry costs, while maintaining the stability and availability of more expensive options.

Quickwit with its original and highly efficient architecture proved to be the ideal candidate to be paired with our OLAP database Clickhouse to run Search + OLAP workloads. During our testing, we have observed that Quickwit could sustain our production workload with efficiency and reliability.

At Etsy, Tantivy, combined with Solr and Etsy's graph retrieval engine, XWalk, fuels our retrieval. Tantivy's fast performance and high code quality allowed us to build a comprehensive retrieval engine, seamlessly integrating filtering, vector search, and full-text search. Its foundation in Rust, an open-source language known for sustainability and security, further solidified our choice.

At Nuclia, we need an alternative to lucene for BM25 search built in Rust and Tantivy is the best match. Tantivy allows us to build a distributed indexing engine with a great performance and clear design. Quickwit's codebase was a great inspiration for building projects.

While analyzing solutions to replace our slow legacy search solution, we selected Tantivy for its high indexing throughput, its low search latency and its high-quality code. Thanks to Tantivy, HumanFirst offers a solution that scales to our customers' needs and allows rapid iteration on their NLU data.

When building bloop's low latency code search engine, we wanted low level control of the search, without having to build a new system from scratch. Using Tantivy, we were able to get a reliable implementation up quickly, freeing up time to invest in other areas - which being in the early stages of product development is the most important thing.

We needed a fast solution in Rust and Python for full-text search that would work well on top of Lance columnar format. We also needed it to be an embedded solution since we're also embedded. Tantivy was the only choice that fit our needs. It was easy to integrate and performed very well. With Tantivy we're able to deliver better retrieval quality to LanceDB users.

ParadeDB selected Tantivy to power its BM25 scoring functionality due to its ergonomic developer interface, Lucene-inspired design, and high-performance search. Thanks to Tantivy, ParadeDB offers BM25 scoring within PostgreSQL, enabling our customers to get high-quality search natively within their database.