classify.py

from regex_processor import classify_with_regex
from bert_processor import classify_with_bert
from llm_processor import classify_with_llm


def classify(logs):
    # process a list of logs and return their labels
    labels = []
    for source, log_msg in logs:
        label = classify_log(source, log_msg)
        labels.append(label)
    return labels


def classify_log(source, log_msg):
    # check source to determine which model to use
    if source == "LegacyCRM":
        # legacy systems use the high-reasoning llm
        label = classify_with_llm(log_msg)
    else:
        # standard systems try regex first for speed
        label = classify_with_regex(log_msg)
        # fallback to bert if regex finds no match
        if not label:
            label = classify_with_bert(log_msg)
    return label


# --- api helper functions ---

def process_single_log(source, message):
    # entry point for a single api request
    return classify_log(source, message)


def process_batch_logs(log_data_list):
    # entry point for bulk api requests
    # expects a list of dictionaries: [{"source": "...", "message": "..."}, ...]
    logs = [(item['source'], item['message']) for item in log_data_list]
    return classify(logs)


# --- developer testing area ---

if __name__ == '__main__':
    # future developers can uncomment the lines below to test logic without an api call

    # test_data = [("ModernCRM", "unauthorized access attempt"), ("LegacyCRM", "retired module warning")]
    # results = classify(test_data)
    # print(f"test results: {results}")
    pass