main.py

from flask import Flask, request, jsonify
from langchain_community.llms import LlamaCpp
from sentence_transformers import SentenceTransformer
from transformers import AutoModel
import torch
from torch.nn.functional import cosine_similarity
import os 

app = Flask(__name__)

n_gpu_layers = 0
n_batch = 1024

llm = LlamaCpp(
    model_path="Phi-3-mini-4k-instruct-q4.gguf",  # path to GGUF file
    temperature=0.1,
    n_gpu_layers=n_gpu_layers,
    n_batch=n_batch,
    verbose=True,
    n_ctx=4096
)
model0 = AutoModel.from_pretrained('sentence-transformers/paraphrase-TinyBERT-L6-v2')

model = SentenceTransformer('sentence-transformers/paraphrase-TinyBERT-L6-v2')

file_size = os.stat('Phi-3-mini-4k-instruct-q4.gguf')
print("model size ====> :", file_size.st_size, "bytes")


@app.route('/cv', methods=['POST'])
def get_skills():
    cv_body = request.json.get('cv_body')

    # Simple inference example
    output = llm(
        f"<|user|>\n{cv_body}<|end|>\n<|assistant|>Can you list the skills mentioned in the CV?<|end|>",
        max_tokens=256,  # Generate up to 256 tokens
        stop=["<|end|>"], 
        echo=True,  # Whether to echo the prompt
    )

    return jsonify({'skills': output})

@app.get('/')
def health():
    return jsonify({'status': 'Worked'})

@app.route('/compare_job', methods=['POST'])
def compare_job():
    job_skills = request.json.get('post')
    employee_skills = request.json.get('job_skills')

    # Validation
    if not isinstance(employee_skills, list) or not all(isinstance(skill, str) for skill in employee_skills):
        raise ValueError("employee_skills must be a list of strings")

    # Encoding skills into embeddings
    job_embedding = model.encode(job_skills)
    employee_embedding = model.encode(employee_skills)

    # Computing cosine similarity between employee skills and the job
    employee_embedding_tensor = torch.from_numpy(employee_embedding).unsqueeze(0)
    job_embedding_tensor = torch.from_numpy(job_embedding).unsqueeze(0)
    similarity_score = cosine_similarity(employee_embedding_tensor, job_embedding_tensor, dim=1)

    return jsonify({"job": job_skills, "similarity_score": similarity_score.item()})



if __name__ == '__main__':
    app.run()