app.py

from pdfminer.high_level import extract_pages
from pdfminer.layout import LTTextContainer
from tqdm import tqdm
import re
import gradio as gr
import os
import accelerate
import spaces
import subprocess
from huggingface_hub import hf_hub_download
from llama_cpp import Llama

from huggingface_hub import login

login(token = os.getenv('HF_TOKEN'))

repo_id = "QuantFactory/Meta-Llama-3-70B-Instruct-GGUF" 
model_id = "Meta-Llama-3-70B-Instruct.Q2_K.gguf"

local_dir = "models"

hf_hub_download(
    repo_id=repo_id,
    filename=model_id,
    local_dir = local_dir
)

def convert_to_json(llm, cv_text, maxtokens, temperature, top_probability):
    json_format = """
            You are an expert at structuring resumes in JSON format. Given a modified resume text, extract the relevant details and convert them into the following structured JSON format:
            
            {
                  "profileDetails": {
                    "firstName": "",
                    "lastName": "",
                    "email": "",
                    "contact": "",
                    "country": "",
                    "jobTitle": "",
                    "social": "",
                    "profileDesc": "",
                    "address": "",
                    "city": "",
                    "state": "",
                    "zipCode": ""
                  },
                  "professionalExperience": [
                    {
                      "positionTitle": "",
                      "location": "",
                      "company": "",
                      "description": "",
                      "startDate": "",
                      "endDate": ""
                    }
                  ],
                  "education": [
                    {
                      "institute": "",
                      "schoolLocation": "",
                      "degree": "",
                      "field": "",
                      "grade": "",
                      "startDate": "",
                      "endDate": ""
                    }
                  ],
                  "skills": [""],
                  "hobbies": [""],
                  "languages": [""],
                  "certifications": [""],
                  "projects": [
                    {
                      "title": "",
                      "description": ""
                    }
                  ],
                  "jobPreferences": {
                    "compTarget": "",
                    "strength": "",
                    "roleTarget": ""
                  },
                  "jobDescription": ""
                }
            
            Instructions:
            - Extract details accurately from the given resume.
            - Ensure proper structuring of dates, responsibilities, and projects.
            - If a field is missing in the input, leave it as an empty string or an empty list where applicable.
            - Maintain proper formatting and avoid unnecessary additions.
            
            Provide the response in a valid JSON format with no additional explanations.
            """
    output = llm.create_chat_completion(
        messages=[
                {"role": "assistant", "content": json_format},
                {
                    "role": "user",
                    "content": cv_text
                }
            ],
        max_tokens=maxtokens,
        temperature=temperature
    )
    output = output['choices'][0]['message']['content']
    return output


def craft_cover_letter(llm, cv_text, job_description, maxtokens, temperature, top_probability):
    instruction = "Given input CV and job description. Please prepare cover letter according to the given job description and give as an output."
    output = llm.create_chat_completion(
            messages=[
                {"role": "assistant", "content": instruction},
                {
                    "role": "user",
                    "content": ' Input CV: ' + cv_text + ' , Job Description: ' + job_description 
                }
            ],
            max_tokens=maxtokens,
            temperature=temperature
        )
    output = output['choices'][0]['message']['content']
        
    return cv_text, output

@spaces.GPU(duration=150)
def pdf_to_text(cv_text, job_description="", function="Convert to JSON", maxtokens=2048, temperature=0, top_probability=0.95):
    llm = Llama(
        model_path="models/" + model_id,
        flash_attn=True,
        n_gpu_layers=81,
        n_batch=1024,
        n_ctx=8192,
    )
    if(function == 'Cover Letter'):
        _, crafted_cv = craft_cover_letter(llm, cv_text, job_description, maxtokens, temperature, top_probability)
    else:
        crafted_cv = convert_to_json(llm, cv_text, maxtokens, temperature, top_probability)
    return crafted_cv

temp_slider = gr.Slider(minimum=0, maximum=2, value=0.9, label="Temperature Value")
prob_slider = gr.Slider(minimum=0, maximum=1, value=0.95, label="Max Probability Value")
max_tokens = gr.Number(value=600, label="Max Tokens")
cv_file = gr.File(label='Upload the CV')
function = gr.Radio(["Convert to JSON", "Cover Letter"])

prompt_text = gr.Textbox(label='Enter the job description')
output_text = gr.Textbox()
iface = gr.Interface(
    fn=pdf_to_text,
    inputs=['text', prompt_text, function],
    outputs=['text'],
    title='Create a Cover Letter or convert PDF to JSON',
    description="This application assists to create a cover letter based on input job description",
    theme=gr.themes.Soft(),
)
iface.launch()