turbopuffer
Vector databaseturbopuffer
Managed vector and full-text search engine storing namespaces on object storage with an NVMe cache.
- Provider
- default
- Kind
- Vector database
- Category
- Data
- Official docs
- https://turbopuffer.com/docs/schema
| property | type | values | required | kind default | this component |
|---|---|---|---|---|---|
| deployment | enum | multi-tenant | single-tenant | byoc | no | multi-tenant | inherited |
| dimensions | integer | — | no | — | inherited |
| distanceFunction | enum | cosine-distance | euclidean-squared | no | — | inherited |
| distanceMetric | enum | cosine | euclidean | dot-product | no | — | inherited |
| hybridSearch | boolean | — | no | false | ▸ true |
| indexType | enum | flat | hnsw | ivf | disk-ann | lsh | no | — | inherited |
| managed | boolean | — | no | false | ▸ true |
| metadataFiltering | boolean | — | no | false | ▸ true |
| namespaces | boolean | — | no | false | ▸ true |
| quantization | enum | none | scalar | product | binary | no | none | inherited |
| replicas | integer | — | no | — | inherited |
| storageBackend | enum | object-storage | no | object-storage | inherited |
Every rule whose selector covers this component's kind or its category. Whether a rule fires depends on the design: its properties, its edges and its groups.
No rule reads this kind yet. That is a gap, not a guarantee.
nodes:
- { id: "turbopuffer", component: "turbopuffer" }Alibaba Cloud Milvusalibaba-milvusAmazon S3 Vectorsaws-s3-vectorsAgent Memorycloudflare-agent-memoryCloudflare Vectorizecloudflare-vectorizeChromachromaDeep LakedeeplakeEpsillaepsillaKDB.AIkdb-aiLanceDBlancedbMarqomarqoMilvusmilvusMyScalemyscalepgvectorpgvectorpgvectorscalepgvectorscalePineconepineconeQdrantqdrantValdvaldGeneric vector databasevector-databaseVespa Cloudvespa-cloudWeaviateweaviateZilliz Cloudzilliz-cloudDigitalOcean Vector Databasesdigitalocean-vector-databaseVertex AI Vector Searchgoogle-vertex-ai-vector-search
# yaml-language-server: $schema=../../schema/component.schema.json # turbopuffer: managed search engine keeping state in object storage with an NVMe and memory cache. # Docs: namespaces are created implicitly on first write, filters run against an inverted index and # BM25 full-text search combines with vector queries; multi-tenant is the default deployment. schemaVersion: "1" type: "component" id: "turbopuffer" kind: "vector-database" provider: "default" docs: "https://turbopuffer.com/docs/schema" defaults: metadataFiltering: true hybridSearch: true namespaces: true managed: true properties: deployment: type: "enum" values: ["multi-tenant", "single-tenant", "byoc"] default: "multi-tenant" storageBackend: type: "enum" values: ["object-storage"] default: "object-storage" distanceFunction: type: "enum" values: ["cosine-distance", "euclidean-squared"]