Skip to content

Workflows Usage Guide 🚀

The run_workflow.py script is a powerful tool to execute sequences of prompts dynamically. This page provides practical usage examples and details about how to run and verify your workflows.

🛠️ Usage Examples

The simulation engine allows you to test your workflows locally. It reads variables from user inputs and outputs mocked data based on the testData within the prompts.

1. Run a Simple Workflow

To run a workflow with a single input via the command line:

python3 tools/scripts/run_workflow.py workflows/technical/agentic_coding.workflow.yaml -i product_concept="A new time-tracking app"

2. Supply Inputs via a File

For workflows requiring many inputs, or when inputs contain sensitive data, use an input file (YAML or JSON):

# Example inputs.yaml file
# product_concept: "A secure messaging platform"
# target_audience: "Enterprise customers"

python3 tools/scripts/run_workflow.py workflows/technical/agentic_coding.workflow.yaml -f inputs.yaml

3. Run in Verbose Mode

To debug a workflow and see the exact inputs being passed to each step during execution, use the -v (verbose) flag:

python3 tools/scripts/run_workflow.py workflows/technical/agentic_coding.workflow.yaml -i product_concept="A new time-tracking app" -v

🔍 How it works

The run_workflow.py script does NOT make actual LLM calls. It utilizes the testData field found within the corresponding .prompt.yaml files.

If a prompt's inputs match the variables specified in a testData entry, the simulator will return that specific expected string. Otherwise, it will return the first available test output, or a generic placeholder.

Example testData matching:

# In your 01_topic_generator.prompt.yaml file:
testData:
  - inputs:
      user_name: "Alice"
    expected: "Hello Admin Alice, welcome to the system."

If you start your workflow with user_name="Alice", the simulator resolves it correctly and passes the "expected" string to the next step.

💡 Best Practices

  • Provide testData: Always populate your prompts with exhaustive testData elements to provide reliable test scenarios.
  • Test Incrementally: Break complex workflows into multiple independent workflows to simplify testing before chaining them.
  • Avoid Command Line Secrets: Do not pass secrets directly via -i. Always rely on an inputs file (-f) for sensitive execution tokens.