brussle (Python)

from brussle import And, Client

db = Client(api_key="...")
ns = db.namespace("acme/prod/tenant_123")
ns.write(upsert=[{"id": "t_123", "attributes": {"plan": "pro"}, "state": {"subject": "Refund"}}])
page = ns.query(filters=And(("attributes.plan", "Eq", "pro"), ("answers.needs_escalation.p", "Gte", 0.85)), top_k=10)

To load many documents, import_documents takes any iterable and writes it in batches, 4 at a time on a thread pool, with retries:

summary = ns.import_documents(records(), concurrency=4)
print(summary.documents, summary.failures)

Methods mirror the API's routes. Every route that changes something takes idempotency_key=. Reads, writes and calls with a key are retried after network errors, timeouts and 5xx, 2 times by default (max_retries). Requests and responses are plain JSON, typed in brussle.types. Filters and rank_by are tuples. Errors raise ApiError with the API's code and details, or unknown for an error response that is not the API's JSON (such as a proxy's error page), or a success whose body is not JSON. A call your organization's plan doesn't include raises plan_required (402), and details["required_plan"] names the plan that does. See the SDK overview.