turbopuffer API - Conare

Conare serves turbopuffer’s API at https://api.conare.ai/tpuf: keep your turbopuffer code and SDK, and point them at Conare with a Conare key.

curl https://api.conare.ai/tpuf/v2/namespaces/products/query \
  -H "Authorization: Bearer $CONARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"rank_by": ["vector", "ANN", [0.1, 0.2, 0.3, 0.4]], "top_k": 10}'

import os
import turbopuffer

tpuf = turbopuffer.Turbopuffer(
    api_key=os.environ["CONARE_API_KEY"],
    region="gcp-us-central1",  # any value: Conare ignores it
    base_url="https://api.conare.ai/tpuf/{region}",
)

ns = tpuf.namespace("products")
ns.write(
    upsert_rows=[{"id": 1, "vector": [0.1, 0.2, 0.3, 0.4], "title": "refunds"}],
    distance_metric="cosine_distance",
)
result = ns.query(rank_by=("vector", "ANN", [0.1, 0.2, 0.3, 0.4]), top_k=10)

Response

{
  "billing": { "billable_logical_bytes_queried": 0, "billable_logical_bytes_returned": 0 },
  "performance": { "cache_hit_ratio": 1.0, "cache_temperature": "hot", "query_execution_ms": 0, "server_total_ms": 0 },
  "rows": [{ "$dist": 0.0, "id": 1 }]
}

You bring your own vectors, and these namespaces are separate from the rest of the API’s.

Supported

Area	Supported
Writes	Upserts, patches, deletes, conditional writes, delete or patch by filter
Queries	Vector (ANN and exact), BM25 full-text, order by attribute, filters, `Count` aggregation, up to 16 queries in one call
Namespaces	List, schema, metadata, cache warming, delete, copy and branch
Distance	`cosine_distance`, `euclidean_squared`
`top_k`	Up to 10,000

Anything else, such as group_by or several vector columns, returns 400 with a message that names it.

Limits

Limit
Queries	1,000 a second per organization
Writes, other calls	100 a second each
Request body	Up to 64 MiB