[FunctionTool] Use docstring_parser to infer description for FunctionTool

llama-index-core/llama_index/core/tools/function_tool.py

@@ -1,12 +1,12 @@
 import asyncio
-from inspect import signature
 from typing import TYPE_CHECKING, Any, Awaitable, Callable, Optional, Type
 
 if TYPE_CHECKING:
  from llama_index.core.bridge.langchain import StructuredTool, Tool
 from llama_index.core.bridge.pydantic import BaseModel
 from llama_index.core.tools.types import AsyncBaseTool, ToolMetadata, ToolOutput
 from llama_index.core.tools.utils import create_schema_from_function
+import docstring_parser
 
 AsyncCallable = Callable[..., Awaitable[Any]]
 
@@ -22,7 +22,8 @@ async def _async_wrapped_fn(*args: Any, **kwargs: Any) -> Any:
 
 
 class FunctionTool(AsyncBaseTool):
- """Function Tool.
+ """
+ Function Tool.
 
  A tool that takes in a function.
 
@@ -54,8 +55,20 @@ def from_defaults(
  ) -> "FunctionTool":
  if tool_metadata is None:
  name = name or fn.__name__
- docstring = fn.__doc__
- description = description or f"{name}{signature(fn)}\n{docstring}"
+ if not description:
+ doc = docstring_parser.parse(fn.__doc__)
+ doc.meta = [
+ i
+ for i in doc.meta
+ if not isinstance(i, docstring_parser.common.DocstringParam)
+ ]
+ description = docstring_parser.compose(
+ doc,
+ docstring_parser.Style.GOOGLE,
+ rendering_style=docstring_parser.RenderingStyle.CLEAN,
+ indent=" ",
+ )
+
  if fn_schema is None:
  fn_schema = create_schema_from_function(
  f"{name}", fn, additional_fields=None

llama-index-core/llama_index/core/tools/types.py

@@ -58,7 +58,8 @@ def get_name(self) -> str:
  "Deprecated in favor of `to_openai_tool`, which should be used instead."
  )
  def to_openai_function(self) -> Dict[str, Any]:
- """Deprecated and replaced by `to_openai_tool`.
+ """
+ Deprecated and replaced by `to_openai_tool`.
  The name and arguments of a function that should be called, as generated by the
  model.
  """

llama-index-core/llama_index/core/tools/utils.py

@@ -1,7 +1,9 @@
 from inspect import signature
-from typing import Any, Callable, List, Optional, Tuple, Type, Union, cast
+from typing import Any, Callable, List, Optional, Tuple, Type, Union, cast, Dict
 
-from llama_index.core.bridge.pydantic import BaseModel, FieldInfo, create_model
+from llama_index.core.bridge.pydantic import BaseModel, FieldInfo, create_model, Field
+
+import docstring_parser
 
 
 def create_schema_from_function(
@@ -12,23 +14,35 @@ def create_schema_from_function(
  ] = None,
 ) -> Type[BaseModel]:
  """Create schema from function."""
- fields = {}
+ fields: Dict[str, FieldInfo] = {}
  params = signature(func).parameters
+ doc = docstring_parser.parse(func.__doc__)
+
+ params_doc = {param.arg_name: param for param in doc.params}
  for param_name in params:
  param_type = params[param_name].annotation
  param_default = params[param_name].default
+ param_desc = (
+ params_doc[param_name].description if param_name in params_doc else None
+ )
 
  if param_type is params[param_name].empty:
  param_type = Any
 
  if param_default is params[param_name].empty:
  # Required field
- fields[param_name] = (param_type, FieldInfo())
+ field_info = Field()
  elif isinstance(param_default, FieldInfo):
- # Field with pydantic.Field as default value
- fields[param_name] = (param_type, param_default)
+ # Field with pydantic.FieldInfo as default value
+ field_info = param_default
  else:
- fields[param_name] = (param_type, FieldInfo(default=param_default))
+ field_info = Field(default=param_default)
+ field_info = cast(FieldInfo, field_info)
+
+ if param_desc:
+ field_info.description = param_desc
+
+ fields[param_name] = (param_type, field_info)
 
  additional_fields = additional_fields or []
  for field_info in additional_fields:

llama-index-core/pyproject.toml

@@ -85,6 +85,7 @@ pillow = ">=9.0.0"
 PyYAML = ">=6.0.1"
 llamaindex-py-client = "^0.1.18"
 wrapt = "*"
+docstring-parser = "^0.16"
 
 [tool.poetry.extras]
 gradientai = [