call-tracer is a library for tracing function calls with depth control and filtering.

pip install call-tracer

(Required: python >= 3.6)

from calltracer import tracer

def add_one(n):
    return n+1

def add_two(n):
    n = add_one(n)
    n = add_one(n)
    return n

@tracer
def main():
    n = add_two(4)

main()

Alternatively, you can use tracer.start() instead of the decorator.

tracer.start()
add_two()
tracer.end()

If you want to use tracer to monitor the internal workings of langchain, you'll encounter a tremendous number of calls and returns, each accompanied by very long file paths. Therefore, you may need to customize the tracer slightly:

from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from calltracer import tracer
tracer.set_max_depth(3)
tracer.set_path_cuts(['python3.9'])
tracer.set_path_filters(['pydantic'])


@tracer
def main():
    prompt = ChatPromptTemplate.from_messages([
        ("system", f"You are a helpful assistant"),
        MessagesPlaceholder(variable_name="chat_history"),
        ("user", "{input}")
    ])

main()

• set_max_depth : Set the maximum depth for tracing. (prints depth <= max_depth)
• set_path_cuts : Specify paths to be shortened in the trace output to simplify and clean up the displayed file paths in the trace logs.
  • For example, if you use trace.set_path_cuts(['PythonProjects']), the path /Users/your_name/PythonProjects/your_project/sample.py will be shortened to your_project/sample.py.
• set_path_filters : Specify paths to be excluded from tracing, preventing the trace function from logging calls and returns for files matching these paths.
• set_show_args : If True, the tracer displays the input arguments of the calls.

• Pull requests are welcome.
  • For major changes, please open an issue first to discuss what you would like to change.
  • Please make sure to update tests as appropriate.

MIT