handler.py

import json
import logging
import datetime
from transformers import AutoModelForCausalLM, AutoTokenizer

# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Load configuration settings from a separate file (config.json)
# Example configuration file:
#{
#  "architectures": [
#    "AceCalisto3"
#  ],
#  "attention_probs_dropout_prob": 0.1,
#  "bos_token_id": 0,
#  "eos_token_id": 2,
#  "hidden_act": "gelu",
#  "hidden_dropout_prob": 0.1,
#  "hidden_size": 1024,
# }
try:
    with open('config.json') as f:
        config = json.load(f)
except FileNotFoundError:
    logger.error("Configuration file 'config.json' not found. Using default settings.")
    config = {
        "model_name": "acecalisto3/InstructiPhi",  # Default model name
        "max_length": 16788,                         # Default max length
        "logging_level": "INFO"                   # Default logging level
    }

# Load model and tokenizer
model_name = config["model_name"]
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Set logging level from configuration
logging.basicConfig(level=config["logging_level"])

def handle_request(event, context):
    """Handles incoming requests to the deployed model.

    Args:
        event: The event data from the deployment platform.
        context: The context data from the deployment platform.

    Returns:
        A dictionary containing the response status code and body.
    """

    try:
        # Extract input text from the event
        input_text = event.get('body')
        if not input_text:
            return {
                'statusCode': 400,
                'body': json.dumps({'error': 'Missing input text'})
            }

        # Input validation: Check length 
        if len(input_text) > 1000:  # Set a reasonable limit
            return {
                'statusCode': 400,
                'body': json.dumps({'error': 'Input text is too long'})
            }

        # Tokenize the input text
        input_ids = tokenizer(input_text, return_tensors="pt").input_ids

        # Generate the response using the model
        output = model.generate(input_ids, max_length=config["max_length"])

        # Decode the generated response
        generated_text = tokenizer.decode(output[0], skip_special_tokens=True)

        # Return a successful response with structured output
        return {
            'statusCode': 200,
            'body': json.dumps({
                'response': generated_text,
                'model': model_name,  # Include model name in the output
                'timestamp': datetime.datetime.now().isoformat() 
            })
        }

    except Exception as e:
        # Log the error with more context
        logger.error(f"Error processing request: {e}, input: {input_text}")
        return {
            'statusCode': 500,
            'body': json.dumps({'error': 'Internal server error'})
        }