Debug Recorder

Manual Recording

Working with intelligent systems creates a lot of information for simple transactions.

In order to know what is going on we are going to use the Recorder class to look inside any process methods called.

from dandy.llm import LlmBot
from dandy.recorder import Recorder

Recorder.start_recording(recording_name='tutorial')

canada_capital_intel = LlmBot.process(prompt='Please tell me just the name only of the city that is the capital of Canada?')

capital_description_intel = LlmBot.process(prompt=f'Please describe the following city: {canada_capital_intel.text}')

Recorder.stop_recording('tutorial')

Recorder.to_html_file('tutorial')

Output: Debug HTML File

Note

The Recorder class can debug multiple process sets at the same time by specifying the debugger_name argument.

Decorator Recording

We can accomplish the same as above by using the @recorder_to_html decorator.

from dandy.llm import LlmBot
from dandy.recorder import recorder_to_html

@recorder_to_html(recording_name='tutorial')
def get_canada_capital_description():
    canada_capital_intel = LlmBot.process(prompt='Please tell me just the name only of the city that is the capital of Canada?')
    return LlmBot.process(prompt=f'Please describe the following city: {canada_capital_intel.text}')

capital_description_intel = get_canada_capital_description()

Output: Debug HTML File

Tip

Having the HTML output file open in a window beside your code editor will help you understand what is going on and drastically speed up the development process.