Feedback on Coding Logic [closed]

New to Coding – Need Feedback on My Approach I’m new to coding and software development, and I’m working on a project in Python 3.11 that uses Ultravox (a voice AI) to make outbound calls and collect information about cars. (This is a proxy example for privacy reasons.) I am only posting this to know if my logic is sound or if there's a better to solve this issue. I am hoping to solve my data collection / check list issue via built in coding since I find prompting to be unpredictable / unreliable. How the App Works - The calls collect details about a car, including: Manufacturer (Toyota, Ford, BMW) Model (Camry, F-150, Tesla) Year (2020, 2022, 2023) Engine Type (Gasoline, Hybrid, Electric) Transmission (Automatic, Manual,Single-Speed) Drive Train (4x4, AWD, RWD, FWD) Problem with My Initial Approach - I originally used Regex to scan the transcript and check if all required details were collected before ending the call. The output looked like this: Manufacturer: Tesla Model: Plaid Year: Engine Type: Electric Transmission: Drive Train: Dual Motor Since Year and Transmission were missing, the AI would know to ask about them. Issues with this approach: Regex checking caused a 2-second delay, making the call feel unnatural. Some key details were still missing despite being coded to ask for them. New Idea: Boolean Checklist - Instead of running Regex mid-call, I’m thinking of using a simple True/False checklist to track whether a category was collected. Then, I’d run Regex only at the end to verify completeness. Example Boolean Checklist: Manufacturer: True Model: True Year: False Engine Type: True Transmission: False Drive Train: True Concerns & Questions Is this an effective way to track missing data in a real-time call scenario? How can I ensure similar categories aren’t conflated? Example: If the car is Electric, we still need to ask if it’s Single or Dual Motor. Just because it’s Electric doesn’t mean it’s a Tesla. The app sometimes forgets a related subcategory. Example: It asks if the car is Electric and if it has a Single-Speed Transmission, but it forgets to ask if it’s Dual Motor. It sometimes skips completely unrelated categories. Example: It gathers all powertrain info but forgets to ask about leather seats. Would a Boolean Checklist help prevent this? (Code Example Below) # Check if all required information has been collected. # async def check_(transcript: str, tool_state: dict = None) -> tuple[bool, list[str]]: """ Check if all required information has been collected. This function serves as a dual validation layer, combining both tool tracking and transcript analysis to ensure complete and accurate information collection. Args: transcript: The call transcript tool_state: State from the track_info tool containing structured data Returns: tuple[bool, list[str]]: (is_complete, missing_items) """ # Patterns to validate information in transcript validation_patterns = { REGEX PATTERN } # Initialize missing items list missing_items = [] # Check tool state first if available if tool_state: for field in validation_patterns.keys(): if field not in tool_state or not tool_state[field]: missing_items.append(field) # Then check transcript for each item transcript = transcript.lower() for field, pattern in validation_patterns.items(): if not re.search(pattern, transcript, re.IGNORECASE): if field not in missing_items: # Don't add duplicates missing_items.append(field) # If an item is found in either tool_state or transcript, remove it from missing_items if tool_state: missing_items = [item for item in missing_items if item not in tool_state or not re.search(validation_patterns[item], transcript, re.IGNORECASE)] return (len(missing_items) == 0, missing_items)

# Check if all required information has been collected.
#
async def check_(transcript: str, tool_state: dict = None) -> tuple[bool, list[str]]:
    """
    Check if all required information has been collected.
    This function serves as a dual validation layer, combining both tool tracking
    and transcript analysis to ensure complete and accurate information collection.

Args:
    transcript: The call transcript
    tool_state: State from the track_info tool containing structured data
    
Returns:
    tuple[bool, list[str]]: (is_complete, missing_items)
"""
# Patterns to validate information in transcript
validation_patterns = {
    REGEX PATTERN
}

# Initialize missing items list
missing_items = []

# Check tool state first if available
if tool_state:
    for field in validation_patterns.keys():
        if field not in tool_state or not tool_state[field]:
            missing_items.append(field)

# Then check transcript for each item
transcript = transcript.lower()
for field, pattern in validation_patterns.items():
    if not re.search(pattern, transcript, re.IGNORECASE):
        if field not in missing_items:  # Don't add duplicates
            missing_items.append(field)

# If an item is found in either tool_state or transcript, remove it from missing_items
if tool_state:
    missing_items = [item for item in missing_items 
                    if item not in tool_state or 
                    not re.search(validation_patterns[item], transcript, re.IGNORECASE)]

return (len(missing_items) == 0, missing_items)

Related Posts

How to encapsulate functions inside a library

Feb 12, 2025  0

How to avoid 'Call super' code smell with multiple base...

Feb 11, 2025  0

Is it a good idea to wrap node.js package functions in ...

Feb 21, 2025  0