guardrails_ai.sdk.chat_completions_api¶

Classes¶

`ChatApi`	Namespaced chat API, mirroring the OpenAI `chat` namespace.
`CompletionsApi`	Guarded chat completions, mirroring the OpenAI completions API.
`GuardedChatCompletion`
`GuardedChatCompletionChunk`

Module Contents¶

class guardrails_ai.sdk.chat_completions_api.ChatApi(*, http_client: httpx.AsyncClient, headers: dict[str, str], max_retries: int)¶

Bases: guardrails_ai.sdk.abstract_client.Client

Namespaced chat API, mirroring the OpenAI chat namespace.

Accessed via client.guards.chat.

completions: CompletionsApi¶

headers: dict[str, str]¶

http_client: httpx.AsyncClient¶

max_retries: int¶

class guardrails_ai.sdk.chat_completions_api.CompletionsApi(*, http_client: httpx.AsyncClient, headers: dict[str, str], max_retries: int)¶

Bases: guardrails_ai.sdk.abstract_client.Client

Guarded chat completions, mirroring the OpenAI completions API.

Accessed via client.guards.chat.completions.

async create(guard_id: str, **kwargs: Unpack[openai.types.chat.completion_create_params.CompletionCreateParamsStreaming]) → openai.AsyncStream[GuardedChatCompletionChunk]¶

async create(guard_id: str, **kwargs: Unpack[openai.types.chat.completion_create_params.CompletionCreateParamsNonStreaming]) → GuardedChatCompletion

Create a guarded chat completion.

Proxies the request through the Guardrails API so that the response is validated by the named Guard before being returned to the caller.

Parameters:

guard_id – The unique id of the Guard to apply to the completion.
stream – If True, returns an async stream of GuardedChatCompletionChunk objects. Defaults to False.
**kwargs – Additional keyword arguments forwarded to the OpenAI chat.completions.create call (e.g. model, messages).

Returns:

A GuardedChatCompletion when stream=False, or an AsyncStream[GuardedChatCompletionChunk] when stream=True.

Example:

# Non-streaming
response = await client.guards.chat.completions.create(
    guard_id="xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx",
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}],
)

# Streaming
async for chunk in await client.guards.chat.completions.create(
    guard_id="xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx",
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True,
):
    print(chunk)

headers: dict[str, str]¶

http_client: httpx.AsyncClient¶

max_retries: int¶

class guardrails_ai.sdk.chat_completions_api.GuardedChatCompletion¶

Bases: openai.types.chat.ChatCompletion

guardrails: guardrails_ai.types.ValidationOutcome | None¶

class guardrails_ai.sdk.chat_completions_api.GuardedChatCompletionChunk¶

Bases: openai.types.chat.ChatCompletionChunk

guardrails: guardrails_ai.types.ValidationOutcome | None¶