Update body_analyzer.py

body_analyzer.py

@@ -2,8 +2,8 @@ import requests
 import os
 import re
 
-HF_API_KEY = os.getenv("HF_API_KEY")  # Hugging Face free account
-HF_HEADERS = {"Authorization": f"Bearer {HF_API_KEY}"}
+HF_API_KEY = os.getenv("HF_API_KEY")
+HF_HEADERS = {"Authorization": f"Bearer {HF_API_KEY}"} if HF_API_KEY else {}
 
 MODELS = {
     "ai_detector": "roberta-base-openai-detector",
@@ -11,7 +11,6 @@ MODELS = {
     "spam": "mrm8488/bert-tiny-finetuned-sms-spam-detection",
 }
 
-# Suspicious patterns to look for
 SUSPICIOUS_PATTERNS = [
     r"verify your account",
     r"urgent action",
@@ -30,9 +29,14 @@ SUSPICIOUS_PATTERNS = [
 ]
 
 def query_hf(model, text):
-    url = f"https://api-inference.huggingface.co/models/{model}"
+    if not HF_API_KEY:
+        return None
     try:
-        res = requests.post(url, headers=HF_HEADERS, json={"inputs": text[:1000]})
+        res = requests.post(
+            f"https://api-inference.huggingface.co/models/{model}",
+            headers=HF_HEADERS,
+            json={"inputs": text[:1000]},
+        )
         return res.json()
     except Exception:
         return None
@@ -41,13 +45,17 @@ def analyze_body(text):
     findings = []
     score = 0
     body_lower = text.lower()
+    highlighted_body = text
 
     # --- 1. Suspicious keyword detection ---
     for pattern in SUSPICIOUS_PATTERNS:
         matches = re.findall(pattern, body_lower)
         for match in matches:
             findings.append(f"Suspicious phrase detected: \"{match}\"")
-            score += 20  # weight for suspicious phrase
+            score += 20
+            highlighted_body = re.sub(
+                match, f"<mark>{match}</mark>", highlighted_body, flags=re.IGNORECASE
+            )
 
     # --- 2. AI-generated text detection ---
     result = query_hf(MODELS["ai_detector"], text)
@@ -55,7 +63,6 @@ def analyze_body(text):
         label = result[0]["label"]
         confidence = result[0]["score"]
         findings.append(f"Body: AI Detector → {label} (confidence {confidence:.2f})")
-        # No score impact yet (just informational)
 
     # --- 3. Sentiment analysis ---
     result = query_hf(MODELS["sentiment"], text)
@@ -63,8 +70,8 @@ def analyze_body(text):
         label = result[0]["label"]
         confidence = result[0]["score"]
         findings.append(f"Body: Sentiment → {label} (confidence {confidence:.2f})")
-        if label.lower() in ["negative"]:
-            score += 10  # negative/urgent tone adds risk
+        if label.lower() == "negative":
+            score += 10
 
     # --- 4. Spam vs Ham detection ---
     result = query_hf(MODELS["spam"], text)
@@ -73,9 +80,9 @@ def analyze_body(text):
         confidence = result[0]["score"]
         findings.append(f"Body: Spam Detector → {label} (confidence {confidence:.2f})")
         if label.lower() == "spam":
-            score += 20  # spam classification increases risk
+            score += 20
 
     if not findings:
-        return ["No suspicious content detected in body."], 0
+        return ["No suspicious content detected in body."], 0, text
 
-    return findings, score
+    return findings, score, highlighted_body