flowshow

Just a super thin wrapper for Python tasks that form a flow.

Installation

uv pip install flowshow

Usage

Flowshow provides a @task decorator that helps you track and visualize the execution of your Python functions. Here's how to use it:

import time
import random

from flowshow import task

# Turns a function into a Task, which tracks a bunch of stuff
@task
def my_function(x):
    time.sleep(0.5)
    return x * 2

# Tasks can also be configured to handle retries
@task(retry_on=ValueError, retry_attempts=10)
def might_fail():
    time.sleep(0.5)
    if random.random() < 0.5:
        raise ValueError("oh no, error!")
    return "done"

@task
def main_job():
    print("This output will be captured by the task")
    for i in range(3):
        my_function(10)
        might_fail()
    return "done"

# Run like you might run a normal function
main_job()

Once you run your function you can expect some nice visuals, like this one:

main_job.plot()

You can also inspect the raw data yourself by running:

main_job.last_run.to_dict()

Show the full dictionary.

{
  "task_name": "main_job",
  "start_time": "2025-02-04T21:25:17.045576+00:00",
  "duration": 8.864794875029474,
  "inputs": {},
  "error": None,
  "retry_count": 0,
  "end_time": "2025-02-04T21:25:25.909997+00:00",
  "logs": "This output will be captured by the task\n",
  "output": "done",
  "subtasks": [
    {
      "task_name": "my_function",
      "start_time": "2025-02-04T21:25:17.045786+00:00",
      "duration": 0.5050525842234492,
      "inputs": {
        "arg0": 10
      },
      "error": None,
      "retry_count": 0,
      "end_time": "2025-02-04T21:25:17.550808+00:00",
      "logs": "",
      "output": 20
    },
    {
      "task_name": "might_fail",
      "start_time": "2025-02-04T21:25:17.550853+00:00",
      "duration": 0.5053939162753522,
      "inputs": {},
      "error": None,
      "retry_count": 0,
      "end_time": "2025-02-04T21:25:18.056233+00:00",
      "logs": "",
      "output": "done"
    },
    {
      "task_name": "my_function",
      "start_time": "2025-02-04T21:25:18.056244+00:00",
      "duration": 0.5052881669253111,
      "inputs": {
        "arg0": 10
      },
      "error": None,
      "retry_count": 0,
      "end_time": "2025-02-04T21:25:18.561502+00:00",
      "logs": "",
      "output": 20
    },
    {
      "task_name": "might_fail",
      "start_time": "2025-02-04T21:25:18.561516+00:00",
      "duration": 2.1351009169593453,
      "inputs": {},
      "error": None,
      "retry_count": 0,
      "end_time": "2025-02-04T21:25:20.696477+00:00",
      "logs": "",
      "output": "done"
    },
    {
      "task_name": "my_function",
      "start_time": "2025-02-04T21:25:20.696511+00:00",
      "duration": 0.5026454580947757,
      "inputs": {
        "arg0": 10
      },
      "error": None,
      "retry_count": 0,
      "end_time": "2025-02-04T21:25:21.199158+00:00",
      "logs": "",
      "output": 20
    },
    {
      "task_name": "might_fail",
      "start_time": "2025-02-04T21:25:21.199213+00:00",
      "duration": 4.711003000382334,
      "inputs": {},
      "error": None,
      "retry_count": 0,
      "end_time": "2025-02-04T21:25:25.909979+00:00",
      "logs": "",
      "output": "done"
    }
  ]
}

You can also get a flat representation of the same data in a dataframe via:

main_job.to_dataframe()

This is what it looks like in Marimo when you evaluate this. Note that we also track the logs of the print statements for later inspection.

Multiple runs

If you run the function multiple times you can also inspect multiple runs:

main_job.runs

This can be useful, but most of the times you're probably interested in the last run.