app.py

import gradio as gr
import os
import sys
import uuid
import chromadb
import pandas as pd
from langchain_groq import ChatGroq
from langchain_community.document_loaders import WebBaseLoader
from langchain_core.prompts import PromptTemplate
from langchain_core.output_parsers import JsonOutputParser

# Get API key from Hugging Face Secrets
GROQ_API_KEY = os.environ.get("GROQ_API_KEY")

# --- Initialize Vector Database on Startup ---
# This part is crucial for loading your portfolio data
try:
    df = pd.read_csv("my_portfolio.csv")
except FileNotFoundError:
    raise FileNotFoundError("my_portfolio.csv not found. Please upload it to your Space.")

client = chromadb.PersistentClient('vectorstore')
collection = client.get_or_create_collection(name="portfolio")

if collection.count() != len(df):
    # Re-populate the collection if the data has changed or is empty
    if collection.count() > 0:
        collection.delete(ids=collection.get()['ids'])
    
    for _, row in df.iterrows():
        collection.add(documents=row["Techstack"],
                       metadatas={"links": row["Links"]},
                       ids=[str(uuid.uuid4())])
    print("✅ Vector database populated with portfolio data.")
else:
    print("✅ Vector database already exists.")

def generate_cold_mail(job_url):
    """
    Main function to generate the cold mail content.
    """
    if not GROQ_API_KEY:
        return "❌ Error: Groq API key is not set in Hugging Face secrets. Please add it to your Space settings."
    if not job_url:
        return "❌ Error: Please provide a job description URL."
        
    # --- 1. Validate Groq API Key ---
    try:
        llm = ChatGroq(
            temperature=0,
            groq_api_key=GROQ_API_KEY,
            model_name="llama3-70b-8192"
        )
        llm.invoke("Test LLM connection.")
    except Exception as e:
        return f"❌ Error: Invalid Groq API key or model unavailable. Details: {e}"

    # --- 2. Scrape and Extract Job Information ---
    try:
        loader = WebBaseLoader(job_url)
        page_data = loader.load().pop().page_content
    except Exception as e:
        return f"❌ Error scraping URL. Please check the URL. Error: {e}"

    prompt_extract = PromptTemplate.from_template(
        """### SCRAPED TEXT FROM WEBSITE: {page_data}
        ### INSTRUCTION: Extract the job posting details and return them in JSON format with keys: `role`, `experience`, `skills` and `description`. Only return the valid JSON.
        ### VALID JSON (NO PREAMBLE):"""
    )
    json_parser = JsonOutputParser()
    chain_extract = prompt_extract | llm | json_parser
    job = chain_extract.invoke(input={'page_data': page_data})
    
    # --- 3. Find Relevant Portfolio Links ---
    job_skills = job.get('skills', [])
    relevant_links = collection.query(query_texts=job_skills, n_results=2).get('metadatas', [])

    # --- 4. Generate Cold Email ---
    prompt_email = PromptTemplate.from_template(
        """### JOB DESCRIPTION: {job_description}
        ### INSTRUCTION: You are Mohan, a business development executive at AtliQ. Write a cold email to the client, describing AtliQ's capabilities in fulfilling their needs. Also add the most relevant ones from the following links to showcase Atliq's portfolio: {link_list}
        ### EMAIL (NO PREAMBLE):"""
    )
    chain_email = prompt_email | llm
    email_content = chain_email.invoke({"job_description": str(job), "link_list": relevant_links})

    return email_content.content

# --- Gradio UI ---
iface = gr.Interface(
    fn=generate_cold_mail,
    inputs=[
        gr.Textbox(label="Job Posting URL"),
    ],
    outputs=gr.Textbox(label="Generated Cold Mail"),
    title="📧 AI Cold Mail Generator",
    description="Provide a job description URL to generate a tailored cold email from AtliQ.",
    theme="huggingface"
)

iface.launch()