Update app.py

app.py

@@ -7,10 +7,9 @@ from huggingface_hub import login
 # ✅ Authenticate using HF token stored in secret
 login(token=os.environ.get("HUGGINGFACEHUB_API_TOKEN"))
 
-# ✅ Load Hugging Face summarization/chat model
+# ✅ Use a lightweight instruction-tuned model compatible with CPU
 summarizer = pipeline("text2text-generation", model="declare-lab/flan-alpaca-base")
 
-
 # ✅ Highlight matching and find missing keywords
 from sklearn.feature_extraction.text import ENGLISH_STOP_WORDS
 
@@ -30,37 +29,34 @@ def highlight_keywords(resume_text, matched):
 # 🔍 Use LLM to extract missing keywords contextually from JD
 def extract_missing_keywords_with_llm(job_desc, resume_text):
     prompt = f"""
-You are an expert resume evaluator.
-
-Given the job description and the resume, identify important skills, tools, and concepts in the job description that are missing or weakly represented in the resume.
-
-List only the missing or insufficiently addressed keywords in bullet format.
+Given the following job description and resume, list the important skills, tools, and concepts from the job description that are missing or weakly represented in the resume.
 
 Job Description:
 {job_desc}
 
 Resume:
 {resume_text}
+
+Only list the missing keywords as bullet points.
 """
-    result = summarizer(prompt, max_new_tokens=300, do_sample=True, temperature=0.7)[0]['generated_text']
+    result = summarizer(prompt, max_new_tokens=300, do_sample=True)[0]['generated_text']
     return result.strip()
 
 # 🔍 Prompt for dynamic section classification and feedback
 def build_dynamic_prompt(job_desc, resume_text, analyze_with_jd):
     prompt = f"""
-You are an expert resume analyst.
-Classify the content of the resume into meaningful categories based on the text (e.g., Technical Skills, Soft Skills, Certifications, Projects, Work Experience, Education, Personal Information).
-You do not need a fixed list of section names—choose the most suitable ones based on the content.
-Then:
-- For each section, summarize its contents.
-- If a job description is provided, compare each section against it.
-- Highlight missing or weak areas relevant to the job.
-- Provide smart, actionable suggestions to improve each section.
-- Output in structured Markdown with headings for each section.
+Analyze the resume below and organize it into meaningful categories (e.g., Skills, Education, Work Experience, etc.).
+
+If a job description is provided, compare it against the resume and suggest improvements section by section.
+
+Job Description:
+{job_desc if analyze_with_jd else '[None provided]'}
+
+Resume:
+{resume_text}
+
+Return structured Markdown with headers for each section and improvement suggestions.
 """
-    if analyze_with_jd and job_desc:
-        prompt += f"\nJob Description:\n{job_desc}\n"
-    prompt += f"\nResume:\n{resume_text}"
     return prompt
 
 # 🧠 Function to call Hugging Face model and get structured resume feedback
@@ -71,14 +67,13 @@ def analyze_resume(job_desc, resume_text, analyze_with_jd):
     user_prompt = build_dynamic_prompt(job_desc, resume_text, analyze_with_jd)
 
     try:
-        response = summarizer(user_prompt, max_new_tokens=768, do_sample=True, temperature=0.7)[0]['generated_text']
-        cleaned = re.sub(rf".*?{re.escape(resume_text)}", "", response, flags=re.DOTALL).strip()
+        response = summarizer(user_prompt, max_new_tokens=512, do_sample=True)[0]['generated_text']
         if analyze_with_jd and job_desc:
             matched, _ = compare_keywords(resume_text, job_desc)
             highlighted_resume = highlight_keywords(resume_text, matched)
             llm_missing_keywords = extract_missing_keywords_with_llm(job_desc, resume_text)
-            return f"### 🔍 Resume with Highlighted Matches\n\n{highlighted_resume}\n\n---\n**✅ Matched Keywords:** {', '.join(sorted(matched)) or 'None'}\n\n**❌ LLM-Inferred Missing Keywords:**\n{llm_missing_keywords}\n\n---\n{cleaned}"
-        return cleaned
+            return f"### 🔍 Resume with Highlighted Matches\n\n{highlighted_resume}\n\n---\n**✅ Matched Keywords:** {', '.join(sorted(matched)) or 'None'}\n\n**❌ LLM-Inferred Missing Keywords:**\n{llm_missing_keywords}\n\n---\n{response.strip()}"
+        return response.strip()
     except Exception as e:
         return f"❌ Error: {str(e)}"