strands.models.litellm

LiteLLM model provider.

• Docs: https://docs.litellm.ai/

LiteLLMModel

class LiteLLMModel(OpenAIModel)

Defined in: src/strands/models/litellm.py:35

LiteLLM model provider implementation.

LiteLLMConfig

class LiteLLMConfig(TypedDict)

Defined in: src/strands/models/litellm.py:38

Configuration options for LiteLLM models.

Attributes:

• model_id - Model ID (e.g., “openai/gpt-4o”, “anthropic/claude-3-sonnet”). For a complete list of supported models, see https://docs.litellm.ai/docs/providers.
• params - Model parameters (e.g., max_tokens). For a complete list of supported parameters, see https://docs.litellm.ai/docs/completion/input#input-params-1.

__init__

def __init__(client_args: dict[str, Any] | None = None,
             **model_config: Unpack[LiteLLMConfig]) -> None

Defined in: src/strands/models/litellm.py:52

Initialize provider instance.

Arguments:

• client_args - Arguments for the LiteLLM client. For a complete list of supported arguments, see https://github.com/BerriAI/litellm/blob/main/litellm/main.py.
• **model_config - Configuration options for the LiteLLM model.

update_config

@override
def update_config(**model_config: Unpack[LiteLLMConfig]) -> None

Defined in: src/strands/models/litellm.py:69

Update the LiteLLM model configuration with the provided arguments.

Arguments:

• **model_config - Configuration overrides.

get_config

@override
def get_config() -> LiteLLMConfig

Defined in: src/strands/models/litellm.py:80

Get the LiteLLM model configuration.

Returns:

The LiteLLM model configuration.

format_request_message_content

@override
@classmethod
def format_request_message_content(cls, content: ContentBlock,
                                   **kwargs: Any) -> dict[str, Any]

Defined in: src/strands/models/litellm.py:90

Format a LiteLLM content block.

Arguments:

• content - Message content.
• **kwargs - Additional keyword arguments for future extensibility.

Returns:

LiteLLM formatted content block.

Raises:

• TypeError - If the content block type cannot be converted to a LiteLLM-compatible format.

format_request_message_tool_call

@override
@classmethod
def format_request_message_tool_call(cls, tool_use: ToolUse,
                                     **kwargs: Any) -> dict[str, Any]

Defined in: src/strands/models/litellm.py:123

Format a LiteLLM compatible tool call, encoding thought signatures into the tool call ID.

Gemini thinking models attach a thought_signature to each function call. LiteLLM’s OpenAI-compatible interface embeds this signature inside the tool call ID using the __thought__ separator. When reasoningSignature is present and the tool call ID does not already contain the separator, this method encodes it so LiteLLM can reconstruct the Gemini-native format on the next request.

Arguments:

• tool_use - Tool use requested by the model.
• **kwargs - Additional keyword arguments for future extensibility.

Returns:

LiteLLM compatible tool call dict with thought signature encoded in the ID when present.

format_request_messages

@override
@classmethod
def format_request_messages(cls,
                            messages: Messages,
                            system_prompt: str | None = None,
                            *,
                            system_prompt_content: list[SystemContentBlock]
                            | None = None,
                            **kwargs: Any) -> list[dict[str, Any]]

Defined in: src/strands/models/litellm.py:234

Format a LiteLLM compatible messages array with cache point support.

Arguments:

• messages - List of message objects to be processed by the model.
• system_prompt - System prompt to provide context to the model (for legacy compatibility).
• system_prompt_content - System prompt content blocks to provide context to the model.
• **kwargs - Additional keyword arguments for future extensibility.

Returns:

A LiteLLM compatible messages array.

format_chunk

@override
def format_chunk(event: dict[str, Any], **kwargs: Any) -> StreamEvent

Defined in: src/strands/models/litellm.py:259

Format a LiteLLM response event into a standardized message chunk.

Extends OpenAI’s format_chunk to:

1. Handle metadata with prompt caching support.
2. Extract thought signatures that LiteLLM embeds in tool call IDs for Gemini thinking models.

Arguments:

• event - A response event from the LiteLLM model.
• **kwargs - Additional keyword arguments for future extensibility.

Returns:

The formatted chunk.

Raises:

• RuntimeError - If chunk_type is not recognized.

stream

@override
async def stream(messages: Messages,
                 tool_specs: list[ToolSpec] | None = None,
                 system_prompt: str | None = None,
                 *,
                 tool_choice: ToolChoice | None = None,
                 system_prompt_content: list[SystemContentBlock] | None = None,
                 **kwargs: Any) -> AsyncGenerator[StreamEvent, None]

Defined in: src/strands/models/litellm.py:315

Stream conversation with the LiteLLM model.

Arguments:

• messages - List of message objects to be processed by the model.
• tool_specs - List of tool specifications to make available to the model.
• system_prompt - System prompt to provide context to the model.
• tool_choice - Selection strategy for tool invocation.
• system_prompt_content - System prompt content blocks to provide context to the model.
• **kwargs - Additional keyword arguments for future extensibility.

Yields:

Formatted message chunks from the model.

structured_output

@override
async def structured_output(
        output_model: type[T],
        prompt: Messages,
        system_prompt: str | None = None,
        **kwargs: Any) -> AsyncGenerator[dict[str, T | Any], None]

Defined in: src/strands/models/litellm.py:369

Get structured output from the model.

Some models do not support native structured output via response_format. In cases of proxies, we may not have a way to determine support, so we fallback to using tool calling to achieve structured output.

Arguments:

• output_model - The output model to use for the agent.
• prompt - The prompt messages to use for the agent.
• system_prompt - System prompt to provide context to the model.
• **kwargs - Additional keyword arguments for future extensibility.

Yields:

Model events with the last being the structured output.