functions.py

def generate_text(prompt, model="gpt-4o-mini", system_prompt="You are a helpful assistant.", 
                  temperature=0.7, max_tokens=4500, max_retries=4, base_delay=10):
    """
    Generates text using litellm's completion API with retry handling for rate limit errors.

    Args:
        prompt (str): The user input prompt.
        model (str): The model to use for text generation.
        system_prompt (str): The system message for context.
        temperature (float): Sampling temperature.
        max_tokens (int): Maximum number of tokens to generate.
        max_retries (int): Maximum retry attempts for rate limit errors.
        base_delay (int): Base delay in seconds for exponential backoff.

    Returns:
        str or None: The generated text, or None if retries fail.
    """
    messages = [
        {"role": "system", "content": system_prompt},
        {"role": "user", "content": prompt}
    ]

    retries = 0

    while retries <= max_retries:
        try:
            # Make the API call
            response = completion(
                model=model,
                messages=messages,
                temperature=temperature,
                max_tokens=max_tokens
            )
            # Return the generated content
            return response.choices[0].message.content.strip()

        except RateLimitError as e:
            # Retry logic for rate limit errors
            print("RateLimitError")
            retries += 1
            wait_time = base_delay * (2 ** (retries - 1))  # Exponential backoff
            print(f"Rate limit hit. Retry {retries}/{max_retries} in {wait_time} seconds...")
            time.sleep(wait_time)

        except Exception as e:
            # Handle other types of exceptions
            print(f"An error occurred: {e}")
                        # Retry logic for rate limit errors
            retries += 1
            wait_time = base_delay * (2 ** (retries - 1))  # Exponential backoff
            print(f"Rate limit hit. Retry {retries}/{max_retries} in {wait_time} seconds...")
            time.sleep(wait_time)
        

 

import mimetypes

def transcribe_with_deepgram(api_key, audio_path):
    endpoint = "https://api.deepgram.com/v1/listen?model=nova-3&smart_format=true"
    mime_type, _ = mimetypes.guess_type(audio_path)
    if mime_type is None:
        mime_type = "audio/wav"  # default fallback
    
    headers = {
        "Authorization": f"Token {api_key}",
        "Content-Type": mime_type
    }

    with open(audio_path, "rb") as f:
        response = requests.post(endpoint, headers=headers, data=f)

    response.raise_for_status()
    return response.json()