Skip to content

API reference

The public surface (sweep_agent.__all__): the agent loop, the LLM backend interface, and the tool registry. Auto-generated from source docstrings.

Agent

sweep_agent.Agent 

Agent(llm: BaseLLM, tools: Registry = <factory>, system_prompt: str = <factory>, max_steps: int = 12, history: list[ChatMessage] = <factory>, tool_selector: Callable[[str, list[str]], list[str]] | None = None, _used_tools: set[str] = <factory>, history_char_budget: int = 56000, last_tool_specs: list[dict] | None = None)

Agent(llm: 'BaseLLM', tools: 'Registry' = , system_prompt: 'str' = , max_steps: 'int' = 12, history: 'list[ChatMessage]' = , tool_selector: 'Callable[[str, list[str]], list[str]] | None' = None, _used_tools: 'set[str]' = , history_char_budget: 'int' = 56000, last_tool_specs: 'list[dict] | None' = None)

history_char_budget class-attribute 

history_char_budget = 56000

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

int('0b100', base=0) 4

max_steps class-attribute 

max_steps = 12

int([x]) -> integer int(x, base=10) -> integer

Convert a number or string to an integer, or return 0 if no arguments are given. If x is a number, return x.int(). For floating point numbers, this truncates towards zero.

If x is not a number or if base is given, then x must be a string, bytes, or bytearray instance representing an integer literal in the given base. The literal can be preceded by '+' or '-' and be surrounded by whitespace. The base defaults to 10. Valid bases are 0 and 2-36. Base 0 means to interpret the base from the string as an integer literal.

int('0b100', base=0) 4

chat 

chat(user_message: str) -> ChatMessage

Run one user→assistant turn (executing any intermediate tool calls).

iter_chat 

iter_chat(user_message: str) -> Iterator[AgentStep]

Stream each step (tool call + result, or final reply) of one turn.

LLM backend

Subclass this to point the agent at any endpoint.

sweep_agent.BaseLLM

Bases: abc.ABC

Minimal protocol every backend must implement.

chat 

chat(
    messages: list[ChatMessage],
    tools: Iterable[dict[str, Any]] | None = None,
    *,
    temperature: float = 0.2,
    max_tokens: int | None = None,
    **kwargs: Any
) -> ChatMessage

Send one round; return the assistant message (may include tool calls).

Tools

Every registered tool is a Tool; call it directly via .fn(params) (no LLM):

from sweep_agent.tools.inspect import inspect_file, InspectFileParams
inspect_file.fn(InspectFileParams(path="vp_init.npy"))

sweep_agent.Tool 

Tool(
    name: str,
    description: str,
    params_model: Type[BaseModel],
    fn: Callable[[BaseModel], Any],
)

Tool(name: 'str', description: 'str', params_model: 'Type[BaseModel]', fn: 'Callable[[BaseModel], Any]')