core.response_options

Frozen dataclass grouping optional parameters for generate_response.

class ResponseOptions(model=None, system_instructions=None, response_format=None, response_model=None, condition=None, abort_condition=None, strict=False, history=None, dependencies=None)[source]

Bases: object

Configuration for a single generate_response call.

All fields are optional. Provider-specific parameters (temperature, max_tokens, tools, tool_choice) are NOT here – they stay as **kwargs on FFAI.generate_response().

Parameters:

model (str | None)
system_instructions (str | None)
response_format (str | dict | None)
response_model (type | None)
condition (str | None)
abort_condition (str | None)
strict (bool)
history (list[str] | None)
dependencies (list[str] | None)

model

Override model for this call.

Type:: str | None

system_instructions

Override system instructions for this call.

Type:: str | None

response_format

Response format hint (e.g. {"type": "json_object"}).

Type:: str | dict | None

response_model

Pydantic BaseModel subclass for structured output.

Type:: type | None

condition

Expression evaluated before execution. If false, prompt is skipped with status="skipped".

Type:: str | None

abort_condition

Stored for future DAG executor use.

Type:: str | None

strict

If True, raise ValueError on unknown {{name.response}} references.

Type:: bool

history

List of prompt names to include in conversation context.

Type:: list[str] | None

dependencies

Prompt names this call depends on.

Type:: list[str] | None

model: str | None = None

system_instructions: str | None = None

response_format: str | dict | None = None

response_model: type | None = None

condition: str | None = None

abort_condition: str | None = None

strict: bool = False

history: list[str] | None = None

dependencies: list[str] | None = None

classmethod from_dict(d)[source]

Construct from an execute_graph prompt dict.

Known option keys are mapped to fields. prompt, prompt_name, and sequence are silently ignored. All other unknown keys are silently ignored.

Parameters:: d (dict[str, Any]) – Dict with keys like prompt_name, prompt, model, condition, etc.
Returns:: ResponseOptions with matched fields set.
Return type:: ResponseOptions

Classes