Update app.py

app.py

@@ -1,10 +1,12 @@
+# Update imports
 import streamlit as st
 import requests
 import json
 import webbrowser
 from io import StringIO
-from groq import Groq
 from bs4 import BeautifulSoup
+import google.generativeai as genai
+import os
 
 # Initialize session state
 if 'original_resume' not in st.session_state:
@@ -20,52 +22,74 @@ def scrape_website(url):
     soup = BeautifulSoup(response.text, 'html.parser')
     return soup.get_text()
 
-def extract_keywords(job_description, client):
-    completion = client.chat.completions.create(
-        model="llama-3.1-70b-versatile",
-        messages=[
-            {
-                "role": "system",
-                "content": (
-                    "You are an expert in extracting essential information from job postings for optimal ATS compatibility. "
-                    "Focus on identifying keywords and skills, prioritized by importance."
-                )
-            },
-            {
-                "role": "user",
-                "content": (
-                    f"Extract keywords from this job posting and categorize them by importance. "
-                    f"Return as JSON with exactly these keys: 'high', 'medium', and 'low' containing arrays of strings.\n\n{job_description}"
-                )
-            }
-        ],
-        temperature=1,
-        max_tokens=4096,
-        response_format={"type": "json_object"}
-    )
-    return json.loads(completion.choices[0].message.content)
-
-def adapt_resume(resume_data, keywords, job_description, client):
-    completion = client.chat.completions.create(
-        model="llama-3.1-8b-instant",
-        messages=[
-            {
-                "role": "system",
-                "content": (
-                    "You are a CV coach skilled in resume customization and JSON formatting. "
-                    "Tailor the resume to emphasize relevant keywords while maintaining factual accuracy."
-                )
-            },
-            {
-                "role": "user",
-                "content": f"Keywords: {json.dumps(keywords)}\nResume: {json.dumps(resume_data)}\nJob Description: {job_description}"
-            }
-        ],
-        temperature=0.9,
-        max_tokens=8000,
-        response_format={"type": "json_object"}
+def trim_text(text, max_length=3000):
+    """Trim text while preserving important content"""
+    if len(text) <= max_length:
+        return text
+    # Keep first and last parts
+    half_length = max_length // 2
+    return text[:half_length] + "..." + text[-half_length:]
+
+# Configure Gemini
+def init_gemini(api_key):
+    genai.configure(api_key=api_key)
+    generation_config = {
+        "temperature": 0.7,
+        "top_p": 0.95,
+        "top_k": 40,
+        "max_output_tokens": 8192,
+        "response_mime_type": "application/json",
+    }
+    return genai.GenerativeModel(
+        model_name="gemini-1.5-flash-002",
+        generation_config=generation_config
     )
-    return json.loads(completion.choices[0].message.content)
+
+def extract_keywords(job_description, model):
+    prompt = f"""Extract key ATS keywords from job posting. Return JSON with 'high', 'medium', 'low' arrays:
+    {job_description}"""
+    
+    response = model.generate_content(prompt)
+    return json.loads(response.text)
+
+def validate_resume_schema(resume_data, original_schema):
+    """Validate and enforce resume schema consistency"""
+    def get_schema_structure(data):
+        if isinstance(data, dict):
+            return {k: get_schema_structure(v) for k, v in data.items()}
+        elif isinstance(data, list):
+            return [get_schema_structure(data[0])] if data else []
+        else:
+            return type(data).__name__
+    
+    original_structure = get_schema_structure(original_schema)
+    new_structure = get_schema_structure(resume_data)
+    return original_structure == new_structure
+
+def adapt_resume(resume_data, keywords, job_description, model, max_retries=3):
+    original_schema = resume_data.copy()
+    
+    for attempt in range(max_retries):
+        try:
+            prompt = f"""You are a CV coach skilled in resume customization and JSON formatting. Consider what the ideal candidate for the target role would be like. You have creative freedom to tailor the provided base Original CV JSON to align with the company's needs using relevant keywords based on their importance level (high: 3x, medium: 2x, low: 1x). Use the language established in the Keywords. IMPORTANT: Consider the company's research and job details to enhance the CV's relevance.
+            Embed relevant keywords throughout specific sections: 'summary', 'experience', 'volunteer', 'interests', 'awards', 'projects' and 'skills' without altering any factual information.
+            Output the modified CV as JSON, strictly maintaining the original structure, keys, and logical flow of information. Use the language specified in the Keywords and avoid adding any fabricated details.
+            Retain specific keys such as 'id' and 'url' as in the original CV, as well as any key found within the 'metadata' section of the schema.
+            Schema: {json.dumps(original_schema)}
+            Keywords: {json.dumps(keywords)}
+            Job: {job_description}"""
+            
+            response = model.generate_content(prompt)
+            tailored_resume = json.loads(response.text)
+            
+            if validate_resume_schema(tailored_resume, original_schema):
+                return tailored_resume
+                
+        except Exception as e:
+            if attempt == max_retries - 1:
+                raise e
+            
+    raise ValueError("Schema validation failed")
 
 def calculate_resume_match(resume_data, keywords):
     """Calculate match score between resume and keywords"""
@@ -105,16 +129,13 @@ st.set_page_config(page_title="Resume Tailor", page_icon="📄", layout="wide")
 # Header
 st.title("🎯 AI Resume Tailor")
 st.markdown("### Transform your resume for your dream job")
-
 # Sidebar with API key
 with st.sidebar:
-    api_key = st.text_input(
-        "Groq API Key", 
-        type="password",
-        help="Get your API key at https://console.groq.com/keys"
-    )
+    st.markdown("### Google API Key")
+    st.markdown("This tool works with Google's Gemini model, which you can use for free. For more information, visit [Google AI Studio](https://ai.google.dev/aistudio).")
+    api_key = st.secrets["google_api_key"]
     if not api_key:
-        st.markdown("[Get API Key](https://console.groq.com/keys)")
+        st.error("API key not found in secrets. Please add your API key to the secrets.")
 
 # Main input section
 col1, col2 = st.columns(2)
@@ -131,13 +152,13 @@ if st.button("🚀 Tailor Resume", type="primary", use_container_width=True):
     if job_url and api_key and resume_file:
         try:
             with st.status("🔄 Processing...") as status:
-                # Initialize client
-                client = Groq(api_key=api_key)
+                # Initialize Gemini
+                model = init_gemini(api_key)
                 
-                # Scrape and process
+                # Rest of the processing remains the same, just using model instead of client
                 status.update(label="Analyzing job posting...")
                 job_description = scrape_website(job_url)
-                keywords = extract_keywords(job_description, client)
+                keywords = extract_keywords(job_description, model)
                 st.session_state['keywords'] = keywords
                 
                 status.update(label="Tailoring resume...")
@@ -145,7 +166,7 @@ if st.button("🚀 Tailor Resume", type="primary", use_container_width=True):
                     st.session_state['original_resume'],
                     keywords,
                     job_description,
-                    client
+                    model
                 )
                 st.session_state['tailored_resume'] = tailored_resume
                 status.update(label="✅ Done!", state="complete")