Update app.py

app.py

@@ -31,16 +31,17 @@ def initialize_ner_pipeline():
 # Step 3: Load custom resume generation model (.h5 file)
 def load_resume_model():
     try:
-        return load_model("resume_enhanced_generator_model.keras")
+        return load_model("resume_generator_model.h5")
     except Exception as e:
         print(f"Error loading resume generation model: {e}")
         return None
 
 # Function to extract text from a PDF file
-def extract_text_from_pdf(pdf_file_path):
+def extract_text_from_pdf(pdf_binary_content):
     text = ""
     try:
-        pdf_document = fitz.open(pdf_file_path)
+        # Open the PDF from binary content
+        pdf_document = fitz.open(stream=pdf_binary_content, filetype="pdf")
         for page_num in range(len(pdf_document)):
             page = pdf_document.load_page(page_num)
             text += page.get_text()
@@ -49,83 +50,66 @@ def extract_text_from_pdf(pdf_file_path):
     return text
 
 # Function to extract text from a Word document
-def extract_text_from_word(docx_file_path):
+def extract_text_from_word(docx_binary_content):
     text = ""
     try:
-        doc = Document(docx_file_path)
+        # Open the DOCX file from binary content
+        doc = Document(docx_binary_content)
         for paragraph in doc.paragraphs:
             text += paragraph.text + "\n"
     except Exception as e:
         return f"Error extracting text from Word document: {e}"
     return text
 
-# Function to clean NER results and filter out irrelevant entities
-def clean_ner_results(ner_results):
-    cleaned_entities = []
-    for entity in ner_results:
-        entity_text = entity['word']
-        entity_type = entity['entity']
-        
-        # Filter out short or irrelevant tokens (e.g., single characters, punctuation)
-        if len(entity_text) > 2 and entity_text.isalpha():  # Only keep meaningful words
-            cleaned_entities.append((entity_text, entity_type))
-    
-    return cleaned_entities
-
-# Function to use the custom .h5 resume model for enhancements
-def generate_resume_with_model(resume_text, job_title, resume_model):
+# Function to optimize resume based on job title using the .h5 model
+def enhance_with_custom_model(resume_text, resume_model):
     try:
-        resume_input = tf.convert_to_tensor([resume_text])
-        job_title_input = tf.convert_to_tensor([job_title])
-        
-        enhanced_resume = resume_model.predict([resume_input, job_title_input])
-        return enhanced_resume[0]
+        # Assuming the .h5 model takes text input and generates enhancement suggestions
+        predictions = resume_model.predict([resume_text])
+        enhancements = " ".join(predictions)  # Convert predictions to text (adjust as needed)
+        return enhancements
     except Exception as e:
-        print(f"Error using resume generation model: {e}")
-        return "Error generating enhanced resume using the custom model."
+        return f"Error using the custom .h5 model: {e}"
 
-# Function to parse and enhance resumes using NER, LLaMA, and the custom .h5 model
-def enhance_resume(input_resume, job_title, client, ner_pipeline, resume_model):
+# Function to optimize resume based on job title
+def optimize_resume(resume_text, job_title, client):
+    prompt = f"Optimize the following resume for the job title '{job_title}':\n\n{resume_text}"
+    responses = []
     try:
-        model_enhanced_resume = generate_resume_with_model(input_resume, job_title, resume_model)
-        ner_results = ner_pipeline(input_resume)
-        cleaned_entities = clean_ner_results(ner_results)
-
-        enhanced_text = f"Job Title: {job_title}\nOriginal Resume: {input_resume}\n\nModel Enhancements:\n{model_enhanced_resume}\n\nAdditional Suggestions:\n"
-        for entity_text, entity_type in cleaned_entities:
-            if entity_type == "I-ORG":
-                enhanced_text += f"Consider emphasizing your experience at {entity_text}.\n"
-
-        llama_input = f"Enhance this resume for the job title: {job_title}. Current content: {model_enhanced_resume}"
-        
-        try:
-            llama_response = client.text_generation(prompt=llama_input, max_new_tokens=1000)
-            print("LLaMA Response:", llama_response)  # Debug output
-            llama_enhanced = llama_response.get("generated_text", "No suggestions generated.")
-        except Exception as llama_error:
-            print(f"Error in LLaMA text generation: {llama_error}")
-            llama_enhanced = "Error generating suggestions from LLaMA model. Please try again later."
-
-        return enhanced_text + "\nRefined Suggestions:\n" + llama_enhanced
-
+        for message in client.chat_completion(
+            messages=[{"role": "user", "content": prompt}],
+            max_tokens=1000,
+            stream=True,
+        ):
+            responses.append(message.choices[0].delta.content)
     except Exception as e:
-        print(f"Error in enhance_resume function: {e}")
-        return "Error generating suggestions. Please try again later."
+        return f"Error during model inference: {e}"
+    
+    return ''.join(responses)
 
 # Function to process the resume and job title inputs
 def process_resume(file, job_title, client, resume_model):
     try:
         file_name = file.name
+        file_content = file.read()
+        
         if file_name.endswith(".pdf"):
-            resume_text = extract_text_from_pdf(file.name)
+            # Extract text if the file is a PDF
+            resume_text = extract_text_from_pdf(file_content)
         elif file_name.endswith(".docx"):
-            resume_text = extract_text_from_word(file.name)
+            # Extract text if the file is a Word document
+            resume_text = extract_text_from_word(file_content)
         else:
-            resume_text = file.decode("utf-8")
-
-        enhanced_resume = enhance_resume(resume_text, job_title, client, ner_pipeline, resume_model)
+            # Assume the file is a text file and read it directly
+            resume_text = file_content.decode("utf-8")
+        
+        # Step 1: Use the custom .h5 model to enhance the resume
+        enhanced_resume = enhance_with_custom_model(resume_text, resume_model)
 
-        return enhanced_resume
+        # Step 2: Optimize the enhanced resume using the LLaMA model
+        optimized_resume = optimize_resume(enhanced_resume, job_title, client)
+        
+        return optimized_resume
     except Exception as e:
         return f"Error processing resume: {e}"
 
@@ -141,11 +125,10 @@ interface = gr.Interface(
         gr.File(label="Upload your resume (Word or PDF)"),
         gr.Textbox(lines=1, placeholder="Enter the job title...", label="Job Title"),
     ],
-    outputs=gr.Textbox(label="Enhanced Resume", lines=20),
+    outputs=gr.Textbox(label="Optimized Resume", lines=20),
     title="Resume Enhancement Tool",
     description="Upload your resume and specify a job title to optimize your resume for that position."
 )
 
-
 # Launch the Gradio app
 interface.launch(share=True)