Python API Reference#

This reference is auto-generated from source code docstrings.

Core Classes#

The primary entry points for the NeMo Guardrails library.

`nemoguardrails.rails.llm.llmrails.LLMRails`	Rails based on a given configuration.
`nemoguardrails.rails.llm.config.RailsConfig`	Configuration object for the models and the rails.

Generation Options and Responses#

Control generation behavior and inspect results.

`nemoguardrails.rails.llm.options.GenerationOptions`	A set of options that should be applied during a generation.
`nemoguardrails.rails.llm.options.GenerationResponse`
`nemoguardrails.rails.llm.options.GenerationRailsOptions`	Options for what rails should be used during the generation.
`nemoguardrails.rails.llm.options.GenerationLogOptions`	Options for what should be included in the generation log.
`nemoguardrails.rails.llm.options.GenerationLog`	Contains additional logging information associated with a generation call.
`nemoguardrails.rails.llm.options.GenerationStats`	General stats about the generation.
`nemoguardrails.rails.llm.options.ActivatedRail`	A rail that was activated during the generation.
`nemoguardrails.rails.llm.options.ExecutedAction`	Information about an action that was executed.

Actions#

Register custom Python functions as actions for use in guardrails flows.

nemoguardrails.actions.action

Decorator to mark a function or class as an action.

Enumerations#

`nemoguardrails.rails.llm.options.RailType`	An enumeration.
`nemoguardrails.rails.llm.options.RailStatus`	An enumeration.
`nemoguardrails.rails.llm.options.RailsResult`

Extensibility — LLM types and protocols#

Public types used by the Custom LLM Model and Custom LLM Framework extension paths.

`nemoguardrails.types.LLMModel`	Protocol that all LLM backends must implement.
`nemoguardrails.types.LLMFramework`	Protocol for pluggable LLM framework backends.
`nemoguardrails.types.ChatMessage`	ChatMessage(role: nemoguardrails.types.Role, content: Union[str, List[Dict[str, Any]], NoneType] = None, tool_calls: Optional[List[nemoguardrails.types.ToolCall]] = None, tool_call_id: Optional[str] = None, name: Optional[str] = None, provider_metadata: Dict[str, Any] = <factory>)
`nemoguardrails.types.LLMResponse`	LLMResponse(content: str, reasoning: Optional[str] = None, tool_calls: Optional[List[nemoguardrails.types.ToolCall]] = None, model: Optional[str] = None, finish_reason: Optional[Literal['stop', 'length', 'tool_calls', 'content_filter', 'error', 'other']] = None, stop_sequence: Optional[str] = None, request_id: Optional[str] = None, usage: Optional[nemoguardrails.types.UsageInfo] = None, provider_metadata: Optional[Dict[str, Any]] = None)
`nemoguardrails.types.LLMResponseChunk`	LLMResponseChunk(delta_content: Optional[str] = None, delta_reasoning: Optional[str] = None, delta_tool_calls: Optional[List[nemoguardrails.types.ToolCall]] = None, model: Optional[str] = None, finish_reason: Optional[Literal['stop', 'length', 'tool_calls', 'content_filter', 'error', 'other']] = None, request_id: Optional[str] = None, usage: Optional[nemoguardrails.types.UsageInfo] = None, provider_metadata: Optional[Dict[str, Any]] = None)
`nemoguardrails.types.ToolCall`	ToolCall(id: str, type: str = 'function', function: nemoguardrails.types.ToolCallFunction = <factory>)
`nemoguardrails.types.ToolCallFunction`	ToolCallFunction(name: str, arguments: Dict[str, Any])
`nemoguardrails.types.UsageInfo`	UsageInfo(input_tokens: int = 0, output_tokens: int = 0, total_tokens: int = 0, reasoning_tokens: Optional[int] = None, cached_tokens: Optional[int] = None)