BaseChatOpenAICallOptions

Parameters for audio output. Required when audio output is requested with modalities: ["audio"]. Learn more.

Callbacks for this call and any sub-calls (eg. a Chain calling an LLM). Tags are passed to all callbacks, metadata is passed to handle*Start callbacks.

Runtime values for attributes previously made configurable on this Runnable, or sub-Runnables.

Describes the format of structured outputs. This should be provided if an output is considered to be structured

Maximum number of parallel calls to make.

Overrides the model's configured maxRetries for this call only.

Metadata for this call and any sub-calls (eg. a Chain calling an LLM). Keys should be strings, values should be JSON-serializable.

Output types that you would like the model to generate for this request. Most models are capable of generating text, which is the default:

["text"]

The gpt-4o-audio-preview model can also be used to generate audio. To request that this model generate both text and audio responses, you can use:

["text", "audio"]

Additional options to pass to the underlying axios request.

Version of AIMessage output format to store in message content.

AIMessage.contentBlocks will lazily parse the contents of content into a standard format. This flag can be used to additionally store the standard format as the message content, e.g., for serialization purposes.

• "v0": provider-specific format in content (can lazily parse with .contentBlocks)
• "v1": standardized format in content (consistent with .contentBlocks)

You can also set LC_OUTPUT_VERSION as an environment variable to "v1" to enable this by default.

The model may choose to call multiple functions in a single turn. You can set parallel_tool_calls to false which ensures only one tool is called at most. Learn more

Static predicted output content, such as the content of a text file that is being regenerated. Learn more.

Used by OpenAI to cache responses for similar requests to optimize your cache hit rates. Replaces the user field. Learn more.

Options controlling OpenAI prompt cache behavior.

Used by OpenAI to set cache retention time

Adds a prompt index to prompts passed to the model to track what prompt is being used for a given generation.

Options for reasoning models.

Note that some options, like reasoning summaries, are only available when using the responses API. If these options are set, the responses API will be used to fulfill the request.

These options will be ignored when not using a reasoning model.

Changing this value part-way through a conversation changes a request-level parameter, which invalidates the cached prompt prefix. Models that support it (currently GPT-6) can instead carry the new effort in a configuration_update content block attached to the message that should start using it:

new HumanMessage({
  content: [
    { type: "configuration_update", reasoning: { effort: "high" } },
    { type: "text", text: "Analyze the failure modes." },
  ],
});

The new effort applies from that message onward, until another update overrides it.

Two configuration_update blocks cannot be adjacent in the conversation history, and they cannot be combined with automatic truncation (truncation: "auto") — OpenAI rejects both with an HTTP 400.

Maximum number of times a call can recurse. If not provided, defaults to 25.

An object specifying the format that the model must output.

Unique identifier for the tracer run for this call. If not provided, a new UUID will be generated.

Name for the tracer run for this call. Defaults to the name of the class.

When provided, the completions API will make a best effort to sample deterministically, such that repeated requests with the same seed and parameters should return the same result.

Service tier to use for this request. Can be "auto", "default", or "flex" Specifies the service tier for prioritization and latency optimization.

Abort signal for this call. If provided, the call will be aborted when the signal is aborted.

Stop tokens to use for this call. If not provided, the default stop tokens for the model will be used.

Additional options to pass to streamed completions. If provided, this takes precedence over "streamUsage" set at initialization time.

If true, model output is guaranteed to exactly match the JSON Schema provided in the tool definition. If true, the input schema will also be validated according to https://platform.openai.com/docs/guides/structured-outputs/supported-schemas.

If false, input schema will not be validated and model output will not be validated.

If undefined, strict argument will not be passed to the model.

Tags for this call and any sub-calls (eg. a Chain calling an LLM). You can use these to filter calls.

Timeout for this call in milliseconds.

Specifies which tool the model should use to respond. Can be an OpenAIToolChoice or a ResponsesToolChoice. If not set, the model will decide which tool to use automatically.

A list of tools that the model may use to generate responses. Each tool can be a function, a built-in tool, or a custom tool definition. If not provided, the model will not use any tools.

The verbosity of the model's response.