Tags

Great follow-up! In Metaflow, tags are lightweight metadata labels you can attach to flows, runs, steps, and tasks. They help you organize, filter, search, and track your executions, especially useful in large projects or shared environments.

🏷️ Types of Tags in Metaflow

There are two main categories:

1. System Tags

• Automatically added by Metaflow
• Examples:
  • metaflow_version:2.10.3
  • user:alice
  • runtime:python3.10
  • origin:cli or origin:notebook
• Useful for filtering by user, version, or run environment

2. User Tags

• Tags you manually assign in code or via CLI
• Examples:
  • model:resnet
  • exp:baseline
  • priority:high

🧪 How to Add Tags

🧵 A. From CLI

python my_flow.py run --tag exp:baseline --tag priority:high

These tags are applied to the entire run.

🧵 B. In Code (self.add_tag)

Inside a step:

@step
def start(self):
    self.add_tag("data:normalized")
    self.add_tag("source:csv")
    self.next(self.end)

You can also remove a tag with:

self.remove_tag("data:normalized")

📦 How to Access Tags Programmatically

from metaflow import Flow

run = Flow('MyFlow').latest_run
print("User tags:", run.user_tags)
print("System tags:", run.system_tags)

for step in run:
    for task in step.tasks:
        print("Task tags:", task.tags)

📍 Why Use Tags?

🔍 How to Filter by Tags (CLI)

metaflow --help  # to list supported commands

metaflow status MyFlow --tag exp:baseline

Or programmatically:

from metaflow import Flow

runs = [run for run in Flow('MyFlow').runs() if 'exp:baseline' in run.user_tags]

✅ Summary