Update app.py

app.py

@@ -1,36 +1,50 @@
-
 import pdfplumber
 import gradio as gr
 import torch
 
-from transformers import pipeline
+from transformers import (
+    AutoTokenizer,
+    AutoModelForCausalLM
+)
 
 MODEL_NAME = "Qwen/Qwen2.5-3B-Instruct"
 
 print("Loading model...")
 
-pipe = pipeline(
-    "text-generation",
-    model=MODEL_NAME,
-    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
-    device_map="auto"
+tokenizer = AutoTokenizer.from_pretrained(
+    MODEL_NAME,
+    trust_remote_code=True
 )
 
-print("Model loaded.")
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_NAME,
+    torch_dtype="auto",
+    device_map="auto",
+    trust_remote_code=True
+)
 
+print("Model loaded successfully.")
 
 
 def extract_text(pdf_file):
     text = ""
 
-    with pdfplumber.open(pdf_file.name) as pdf:
-        for page in pdf.pages:
-            page_text = page.extract_text()
-            if page_text:
-                text += page_text + "\n"
+    try:
+        with pdfplumber.open(pdf_file.name) as pdf:
+
+            for page in pdf.pages:
+                page_text = page.extract_text()
+
+                if page_text:
+                    text += page_text + "\n"
+
+    except Exception as e:
+        return f"PDF Extraction Error: {str(e)}"
+
+    print(f"Extracted {len(text)} characters")
 
-    return text[:12000]
-    print(text)
+    # Keep small for free-tier inference
+    return text[:3000]
 
 
 def build_prompt(policy_text):
@@ -38,13 +52,25 @@ def build_prompt(policy_text):
     return f"""
 You are a senior insurance consultant.
 
-Analyze the insurance policy and produce a customer-friendly report.
+Analyze the insurance policy and create a customer-friendly report.
 
-Return your answer in markdown.
+Return markdown.
 
 # Executive Summary
 
-Summarize the policy in simple English.
+Summarize the policy in plain English.
+
+# Customer Risk Score
+
+Rate 1-10 and explain why.
+
+# Policy Complexity Score
+
+Rate 1-10 and explain why.
+
+# Claim Difficulty Score
+
+Rate 1-10 and explain why.
 
 # What Is Covered
 
@@ -58,40 +84,29 @@ Provide bullet points.
 
 Provide bullet points.
 
-# Claim Process
-
-Step-by-step instructions.
-
 # Coverage Gaps
 
-Identify situations where customers may incorrectly assume coverage.
+Identify situations where customers may wrongly assume they are covered.
 
-# Red Flags
+# Claim Checklist
 
-Anything customers commonly overlook.
+Provide step-by-step instructions.
 
-# Questions To Ask Your Insurer
+# Questions To Ask The Insurer
 
 Provide 5 questions.
 
 # Explain Like I'm 15
 
-Explain the policy in simple language.
+Explain the policy simply.
 
-Policy Text:
+POLICY DOCUMENT:
 
 {policy_text}
 """
 
 
-def analyze_policy(pdf_file):
-
-    if pdf_file is None:
-        return "Please upload a policy PDF."
-
-    policy_text = extract_text(pdf_file)
-
-    prompt = build_prompt(policy_text)
+def generate_response(prompt):
 
     messages = [
         {
@@ -104,19 +119,83 @@ def analyze_policy(pdf_file):
         }
     ]
 
-    result = pipe(
+    text = tokenizer.apply_chat_template(
         messages,
-        max_new_tokens=1200,
+        tokenize=False,
+        add_generation_prompt=True
+    )
+
+    inputs = tokenizer(
+        text,
+        return_tensors="pt",
+        truncation=True,
+        max_length=4096
+    )
+
+    inputs = {k: v.to(model.device) for k, v in inputs.items()}
+
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=800,
         temperature=0.2,
-        do_sample=True
+        do_sample=True,
+        pad_token_id=tokenizer.eos_token_id
+    )
+
+    generated_tokens = outputs[0][inputs["input_ids"].shape[1]:]
+
+    response = tokenizer.decode(
+        generated_tokens,
+        skip_special_tokens=True
     )
 
-    return result[0]["generated_text"][-1]["content"]
+    return response
+
+
+def analyze_policy(pdf_file):
+
+    try:
+
+        if pdf_file is None:
+            return "Please upload a policy PDF."
+
+        policy_text = extract_text(pdf_file)
+
+        if len(policy_text.strip()) == 0:
+            return "No text could be extracted from this PDF."
+
+        prompt = build_prompt(policy_text)
+
+        response = generate_response(prompt)
+
+        return response
+
+    except Exception as e:
+
+        print("ERROR:", e)
+
+        return f"""
+# Error
+
+{str(e)}
+"""
+
+
+CUSTOM_CSS = """
+footer {
+    display:none;
+}
+
+.gradio-container {
+    max-width: 1200px !important;
+}
+"""
 
 
 with gr.Blocks(
     title="Insurance Policy Decoder",
-    theme=gr.themes.Soft()
+    theme=gr.themes.Soft(),
+    css=CUSTOM_CSS
 ) as demo:
 
     gr.Markdown(
@@ -125,42 +204,45 @@ with gr.Blocks(
 
 Understand your insurance policy in less than a minute.
 
-Upload a policy PDF to get:
+Upload a policy PDF and receive:
 
-✅ Simple Summary
+✅ Executive Summary
 
 ✅ Coverage Details
 
-✅ Major Exclusions
+✅ Exclusions
 
 ✅ Waiting Periods
 
-✅ Claim Checklist
-
 ✅ Coverage Gaps
 
+✅ Risk Scores
+
+✅ Claim Checklist
+
 ✅ Questions To Ask Your Insurer
 """
     )
 
-    with gr.Row():
-
-        pdf_input = gr.File(
-            label="Upload Insurance Policy",
-            file_types=[".pdf"]
-        )
+    pdf_input = gr.File(
+        label="Upload Insurance Policy PDF",
+        file_types=[".pdf"]
+    )
 
     analyze_btn = gr.Button(
         "Decode Policy",
         variant="primary"
     )
 
-    output = gr.Markdown()
+    output = gr.Markdown(
+        value="Upload a policy document and click **Decode Policy**."
+    )
 
     analyze_btn.click(
         fn=analyze_policy,
         inputs=pdf_input,
-        outputs=output
+        outputs=output,
+        show_progress="full"
     )
 
 demo.launch()