Linear's search & context layer

Their responsiveness and shipping velocity make us feel like we are their only customer.

Tom Moor, Head of Engineering

Why turbopuffer?

Linear was drawn to turbopuffer for its ability to ingest 100s of millions of full-text search documents & vectors without having to think about machine types. Instead, Linear can focus on shipping features.

Linear does hybrid search (FTS + vector) on multiple (org_id, table) namespaces, rank fuse, and re-rank:

                          ┌─turbopuffer queries──┐
                          │  ┌───────────────┐   │░
                         ┌┼─▶│ Issues Vector │──┐│░
                         ││  ├───────────────┤  ││░
                         ├┼─▶│  Issues FTS   │──┤│░
                         ││  ├───────────────┤  ││░
┌────────┐  ┌----------┐ ├┼─▶│Document Vector│──┤│░  ┌------┐   ┌--------┐
│ Linear │  | Cohere   | ││  ├───────────────┤  ││░  | Rank |   | Cohere |
│ Query  │─▶| Embedding|─┼┼─▶│ Document FTS  │──┼┼─▶ | Fuse |──▶| Rerank |
└────────┘  | Model    | ││  ├───────────────┤  ││░  └------┘   └--------┘
            └----------┘ ├┼─▶│Project Vector │──┤│░
                         ││  ├───────────────┤  ││░
                         └┼─▶│  Project FTS  │──┘│░
                          │  └───────────────┘   │░
                          └──────────────────────┘░
                           ░░░░░░░░░░░░░░░░░░░░░░░░

Customer data security is critical to Linear. The ability to use customer managed encryption keys (CMEK) on a per namespace basis stood out.

Results

1. Single system for vector and full-text search
2. Zero-ops search for terabytes of data
3. 70% reduction in cost, unlocking indexing far more data

It's been fun to work with @turbopuffer on this. Admittedly, keyword-based search never really worked great for issues, and the results from embedding+FTS are now actually useful. We'll also use this new infrastructure to power more of Linear's AI functionality

Jori Lallo, Co-founder

turbopuffer in Linear

When a user submits a search, Linear will issue parallel queries across multiple namespaces (documents, issues, projects, comments, attachments, issues, and initiatives) using vector + FTS to return a list of results that then get passed into a reranker:

Linear also leverages turbopuffer for:

1. Finding similar issues when a user opens a new issue
2. Offering suggestions during issue creation, like which user should be assigned or which project the issue should belong to
3. 🔜 Providing context to agentic workflows that are seamlessly integrated into the Linear experience

Linear plans to leverage turbopuffer to power all of their AI features.