Baretools AI

API Reference

← Home  ·  Get Started  ·  Concepts  ·  Providers

---

tool

from baretools import tool

@tool
def my_function(a: int, b: str = "default") -> str:
    """Tool description for the LLM."""
    ...

Marks a function as a tool. Works with both sync and async def functions. Can also be called directly: tool(my_fn).

---

ToolRegistry

from baretools import ToolRegistry

registry = ToolRegistry(
    logger=None,    # logging.Logger | None
    on_event=None,  # Callable[[ToolEvent], None] | None
)

register(fn)

Register a @tool-decorated function.

---

get_schemas(provider="openai", strict=False) → list

Return provider-native schemas for all registered tools.

provider	Return shape
"openai"	[{"type":"function","function":{"name":...,"parameters":...}}]
"anthropic"	[{"name":...,"description":...,"input_schema":...}]
"gemini"	[{"functionDeclarations":[...]}]
"json_schema"	[{"name":...,"parameters":...}]

strict=True applies only to "openai". Returns defensive copies — mutations do not affect the registry.

---

execute(tool_calls, *, parallel=False, max_workers=4, retries=0, retry_delay_seconds=0.0) → list[ToolResult]

Execute a list of ToolCalls synchronously. Errors are captured in ToolResult.error; exceptions never propagate out of execute.

Parameter	Default	Description
parallel	False	Run calls concurrently in a ThreadPoolExecutor
max_workers	4	Thread pool size when parallel=True
retries	0	Additional attempts on exception (0 = no retries)
retry_delay_seconds	0.0	Sleep between retries

---

execute_async(tool_calls, *, parallel=False, max_concurrency=4, retries=0, retry_delay_seconds=0.0) → Coroutine[list[ToolResult]]

Async variant. Drives async def tools natively; wraps sync tools in a thread executor.

---

execute_stream(tool_calls, *, parallel=False, max_workers=4, retries=0, retry_delay_seconds=0.0) → Iterator[ToolResult]

Sync generator. Yields one ToolResult per completed call. With parallel=True, yields in completion order (fastest first). Serial mode preserves input order.

---

execute_stream_async(tool_calls, *, parallel=False, max_concurrency=4, retries=0, retry_delay_seconds=0.0) → AsyncIterator[ToolResult]

Async generator equivalent of execute_stream.

---

parse_tool_calls

from baretools import parse_tool_calls

tool_calls: list[ToolCall] = parse_tool_calls(response, provider="openai")

Parse a provider response into normalised ToolCall TypedDicts. Returns an empty list when the response contains no tool calls.

provider	Expected response type
"openai"	openai.types.chat.ChatCompletionMessage
"anthropic"	anthropic.types.Message
"gemini"	google.genai.types.GenerateContentResponse

---

format_tool_results

from baretools import format_tool_results

messages: list[ProviderToolResult] = format_tool_results(results, provider="openai")

Format list[ToolResult] into provider-native message dicts.

provider	Output	How to use
"openai"	[{"role":"tool","tool_call_id":...,"content":...}]	messages.extend(...)
"anthropic"	[{"type":"tool_result","tool_use_id":...,"content":...}]	{"role":"user","content": messages}
"gemini"	[{"name":...,"response":{...}}]	Part.from_function_response(**item) for each

---

TypedDicts

ToolCall

class ToolCall(TypedDict, total=False):
    id: str | None
    tool_call_id: str | None
    name: str | None
    arguments: dict[str, Any] | str

ToolResult

class ToolResult(TypedDict):
    tool_call_id: str | None
    tool_name: str | None
    output: Any
    error: str | None       # None on success; exception message on failure
    attempts: int           # 1 if no retries configured
    execution_time_ms: int

ToolEvent

class ToolEvent(TypedDict, total=False):
    event: Literal["tool_attempt", "tool_retry", "tool_failed"]
    tool_call_id: str | None
    tool_name: str | None
    attempt: int
    error: str

ProviderToolResult

Union type returned by format_tool_results. Internal structure is provider-specific; pass items directly to your SDK.

---

__version__

import baretools
print(baretools.__version__)  # e.g. "0.3.0"

Read from package metadata at import time. Matches version in pyproject.toml.

This site is open source. Improve this page.