auditqa/utils.py

import re
import logging
import json
from langchain.schema import (
    HumanMessage,
    SystemMessage,
)

def save_logs(scheduler, JSON_DATASET_PATH, logs, feedback=None) -> None:
    """ Every interaction with app saves the log of question and answer, 
        this is to get the usage statistics of app and evaluate model performances.
        Also saves user feedback (when provided).
    """
    if feedback:
        logs["feedback"] = feedback #optional
        
    with scheduler.lock:
        with JSON_DATASET_PATH.open("a") as f:
            json.dump(logs, f)
            f.write("\n")
    print("logging done")

def get_message_template(type, SYSTEM_PROMPT, USER_PROMPT):
    if type == 'NVIDIA':
        messages =  [{"role": "system", "content": SYSTEM_PROMPT},
                {"role":"user","content":USER_PROMPT}]
    elif type == 'DEDICATED':
        messages = [
                 SystemMessage(content=SYSTEM_PROMPT),
                 HumanMessage(content=USER_PROMPT),]
    else:
        messages = None
    
    return messages


def make_html_source(source,i):
    """
    takes the text and converts it into html format for display in "source" side tab
    """
    meta = source.metadata
    content = source.page_content.strip()

    name = meta['subtype']
    card = f"""
        <div class="card" id="doc{i}">
            <div class="card-content">
                <h2>Doc {i} - {meta['subtype']} - Page {int(meta['page'])}</h2>
                <p>{content}</p>
            </div>
            <div class="card-footer">
                <span>{name}</span>
                <a href="{meta['subtype']}#page={int(meta['page'])}" target="_blank" class="pdf-link">
                    <span role="img" aria-label="Open PDF">🔗</span>
                </a>
            </div>
        </div>
        """

    return card


def parse_output_llm_with_sources(output):
    # Split the content into a list of text and "[Doc X]" references
    content_parts = re.split(r'\[(Doc\s?\d+(?:,\s?Doc\s?\d+)*)\]', output)
    parts = []
    for part in content_parts:
        if part.startswith("Doc"):
            subparts = part.split(",")
            subparts = [subpart.lower().replace("doc","").strip() for subpart in subparts]
            subparts = [f"""<a href="#doc{subpart}" class="a-doc-ref" target="_self"><span class='doc-ref'><sup>{subpart}</sup></span></a>""" for subpart in subparts]
            parts.append("".join(subparts))
        else:
            parts.append(part)
    content_parts = "".join(parts)
    return content_parts