dosubot[bot]
commented
1 month ago To pass parameters when using FlowTool to call other programs, you can use either positional or keyword arguments. Here is how you can do it:
Using Positional Arguments
Ensure the number of positional arguments matches the number of inputs expected by the tool. The tool will map these arguments to the corresponding input names.
# Assuming the tool expects two inputs: "input1" and "input2"
result = flow_tool_instance._run("value1", "value2")Using Keyword Arguments
You can pass the parameters as keyword arguments, where the keys are the input names.
# Assuming the tool expects two inputs: "input1" and "input2"
result = flow_tool_instance._run(input1="value1", input2="value2")Code Explanation
The _run method in FlowTool handles the arguments as follows:
- It first retrieves the input names using
get_arg_names(self.inputs). - If the number of positional arguments matches the number of input names, it maps them accordingly.
- If the number of positional arguments does not match, it raises an exception and suggests using keyword arguments instead.
- It then builds a dictionary of tweaks from the arguments and runs the flow with these tweaks.
Here is the relevant part of the code:
def _run(
self,
*args: Any,
**kwargs: Any,
) -> str:
"""Use the tool."""
args_names = get_arg_names(self.inputs)
if len(args_names) == len(args):
kwargs = {arg["arg_name"]: arg_value for arg, arg_value in zip(args_names, args)}
elif len(args_names) != len(args) and len(args) != 0:
raise ToolException(
"Number of arguments does not match the number of inputs. Pass keyword arguments instead."
)
tweaks = {arg["component_name"]: kwargs[arg["arg_name"]] for arg in args_names}
run_outputs = syncify(run_flow, raise_sync_error=False)(
tweaks={key: {"input_value": value} for key, value in tweaks.items()},
flow_id=self.flow_id,
user_id=self.user_id,
)
if not run_outputs:
return "No output"
run_output = run_outputs[0]
data = []
if run_output is not None:
for output in run_output.outputs:
if output:
data.extend(build_data_from_result_data(output, get_final_results_only=self.get_final_results_only))
return format_flow_output_data(data)This method ensures that the parameters are correctly passed and validated before running the flow [1][2].
To continue talking to Dosu, mention @dosu.
vasconceloscezar
Bug Description
I need to use ToolCallAgent components to make a conference reservation system and a station reservation system, and use different processes to complete the two systems respectively. Then I use ToolCallAgent plus Flow as Tool to complete the call, and use AI to determine which system to use. Then when I gave it to Flow as Tool to call, I encountered several problems and hoped to be answered or solved. 1 How to pass parameters when using Flow as Tool to call other programs.
Reproduction
I need to use ToolCallAgent components to make a conference reservation system and a station reservation system, and use different processes to complete the two systems respectively. Then I use ToolCallAgent plus Flow as Tool to complete the call, and use AI to determine which system to use. Then when I gave it to Flow as Tool to call, I encountered several problems and hoped to be answered or solved. 1 How to pass parameters when using Flow as Tool to call other programs.
Expected behavior
I need to use ToolCallAgent components to make a conference reservation system and a station reservation system, and use different processes to complete the two systems respectively. Then I use ToolCallAgent plus Flow as Tool to complete the call, and use AI to determine which system to use. Then when I gave it to Flow as Tool to call, I encountered several problems and hoped to be answered or solved. 1 How to pass parameters when using Flow as Tool to call other programs.
Who can help?
No response
Operating System
windows
Langflow Version
1.0.14
Python Version
None
Screenshot
No response
Flow File
No response