main.py

from flask import Flask, request, jsonify
from langchain_community.llms import LlamaCpp
from sentence_transformers import SentenceTransformer
from transformers import AutoTokenizer, AutoModel
from huggingface_hub import hf_hub_download, HfApi

# cosine_similarity
import torch
from torch.nn.functional import cosine_similarity
import os

app = Flask(__name__)

n_gpu_layers = 0
n_batch = 1024

# تنزيل النموذج باستخدام معالجة الأخطاء
try:
    model_path = hf_hub_download(repo_id="repo_name", filename="model_file_name", force_download=True)
except Exception as e:
    print(f"Error downloading the model: {e}")
    model_path = None

# تأكد من أن النموذج تم تنزيله بنجاح
if model_path:
    llm = LlamaCpp(
        model_path=model_path,  # 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"\n{cv_body}\nCan you list the skills mentioned in the CV?",
        max_tokens=256,  # Generate up to 256 tokens
        stop=[""], 
        echo=True,  # Whether to echo the prompt
    )

    return jsonify({'skills': output})

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

@app.route('/compare', methods=['POST'])
def compare():
    employee_skills = request.json.get('employee_skills')  # string
    jobs_skills = request.json.get('jobs_skills')  # list of strings

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

    job_embeddings = model.encode(jobs_skills)
    employee_embeddings = model.encode(employee_skills)
    sim = []
    employee_embeddings_tensor = torch.from_numpy(employee_embeddings).unsqueeze(0)
    for job_e in job_embeddings:
        job_e_tensor = torch.from_numpy(job_e).unsqueeze(0)
        sim.append(cosine_similarity(employee_embeddings_tensor, job_e_tensor, dim=1))

    max_sim = max(sim)
    index = sim.index(max_sim)
    return jsonify({'job': jobs_skills[index]})

@app.route('/models', methods=['GET'])
def list_models():
    hf_api = HfApi()
    models = hf_api.list_models()
    return jsonify({'models': models})

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