Use AsyncTypeSafeClient to ask questions, list models, and configure asynchronous TypeSafe API requests.
typesafe_sdk.AsyncTypeSafeClient
Create an asynchronous HTTP client for TypeSafe AI API.Explicit options take precedence over environment variables; empty or whitespace-only environment values are ignored.
Logging setupThe SDK logs to the typesafe_sdk logger; configure it through standard logging, or set TYPESAFE_LOG_LEVEL (debug, info, …) for a quick default. Secret headers are redacted from log output; request and response bodies are not.
Parameters:
api_key (str | None, default: None ) –Required API key; may be set via the TYPESAFE_API_KEY environment variable.
model (str | None, default: None ) –Model name; may be set via the TYPESAFE_DEFAULT_MODEL environment variable.
retry (RetryPolicy | None, default: None ) –A RetryPolicy controlling retry behavior; see RetryPolicy for the available options and their defaults. Pass RetryPolicy(max_retries=0) to disable retries.
timeout (float | httpx2.Timeout | None, default: None ) –Timeout for HTTP operations. Inherits http_client.timeout when supplied, otherwise the SDK default.
transport (httpx2.AsyncBaseTransport | None, default: None ) –Optional custom HTTP transport, closed when this SDK client closes.
http_client (httpx2.AsyncClient | None, default: None ) –Optional httpx2.AsyncClient; mutually exclusive with transport. Closed when this SDK client closes.
base_url (str | None, default: None ) –API root; may be set via the TYPESAFE_BASE_URL environment variable.
Raises:
TypeSafeError –The API key is missing or the timeout is invalid.
ValueError –Both transport and http_client are supplied.
Examples:
import asynciofrom typesafe_sdk import AsyncTypeSafeClient, Choice, Noulasync def main() -> None: async with AsyncTypeSafeClient() as client: result = await client.system_one( state="I was charged twice. Please help.", questions={ "billing": Noul(instructions="Is this about billing?"), "tone": Choice( instructions="What is the tone?", criteria={"calm": None, "angry": None}, ), }, ) assert 0 <= result.nouls["billing"].noul <= 1 assert result.choices["tone"].choice in {"calm", "angry"}asyncio.run(main())
models
cachedpropertyAn accessor for the Models API resource.Examples:
async def main() -> None: async with AsyncTypeSafeClient() as client: models = await client.models.list()
system_one
async
Implementation
Overload 1
Overload 2
Answer named questions about text or structured state.See System One for details.Parameters:
state (JSONContent) –Text, a JSON object, or an array to evaluate. See state for details.
questions (Mapping[str, Question]) –Nonempty mapping of names to question objects or raw dictionaries.
model (str | None, default: None ) –Model override; None inherits the client default.
retry (RetryPolicy | None, default: None ) –An optional retry policy to override the client-level value for this call only.
timeout (float | httpx2.Timeout | None, default: None ) –An optional timeout for http operations to override the client-level value for this call only, in seconds.
extra_body (Mapping[str, JSONValue | None] | None, default: None ) –Additional top-level request-body fields, shallow-merged over the body after state, model, and questions are set. Merging is last-write-wins: a key that collides with state, model, or questions overrides it, and object values are replaced rather than deep-merged.
response_model (type[ResponseT] | None, default: None ) –Optional Pydantic BaseModel type describing the JSON response body, including any nested answer models.
Returns:
SystemOneResponse | ResponseT –An instance of response_model, or SystemOneResponse with answers keyed by question
SystemOneResponse | ResponseT –name and model and token usage details when no custom model is supplied.
Raises:
TypeSafeError –Questions are empty or a score question’s criteria list is empty.
TypeSafeAPIError –The server returns an unsuccessful HTTP response after any retries.