Add multilingual support, AI suggestions and Electron desktop app

.gitignore

@@ -1,3 +1,22 @@
+<<<<<<< HEAD
+# Python
+venv/
+spamdetection/venv/
+__pycache__/
+*.pyc
+
+# Environment variables
+.env
+
+# Node
+node_modules/
+
+# Build files
+dist/
+
+# Windows
+*.lnk
+=======
 # --- Python Virtual Environment ---
 # Never push your local environment folder
 venv/
@@ -29,4 +48,5 @@ build/
 # --- VS Code & IDE Settings ---
 .vscode/
 .idea/
-.DS_Store
+.DS_Store
+>>>>>>> upstream/main

backend/api.py

@@ -2,10 +2,34 @@
 import joblib
 import os
 from dotenv import load_dotenv
+from flask_cors import CORS
+from translator import translate_to_english
 
-load_dotenv()
+from pathlib import Path
+import google.generativeai as genai
+import os
+
+load_dotenv(Path(__file__).parent / ".env")
+print("MODEL_PATH =", os.getenv("MODEL_PATH"))
+print("VECTORIZER_PATH =", os.getenv("VECTORIZER_PATH"))
+print("LABEL_ENCODER_PATH =", os.getenv("LABEL_ENCODER_PATH"))
+genai.configure(
+    api_key=os.getenv("GEMINI_API_KEY")
+)
+
+ai_model = genai.GenerativeModel("gemini-2.5-flash")
 
 app = Flask(__name__)
+CORS(app)
+CORS(
+    app,
+    resources={
+        r"/*": {
+            "origins": ["http://localhost:5173"]
+        }
+    }
+)
+
 
 MODEL_PATH = os.getenv("MODEL_PATH")
 VECTORIZER_PATH = os.getenv("VECTORIZER_PATH")
@@ -23,37 +47,92 @@
 def home():
     return "ML API Running 🚀"
 
+def generate_ai_suggestion(text, prediction):
+
+    prompt = f"""
+    Message:
+
+    {text}
+
+    Prediction: {prediction}
+
+    Respond in the same language as the message.
+
+    Give only one short recommendation (under 20 words).
+
+    Examples:
+
+    Safe:
+    This message appears safe.
+
+    Spam:
+    Do not click links or share personal information.
+    """
 
+    try:
+        response = ai_model.generate_content(prompt)
+        return response.text
+
+    except Exception as e:
+        return f"AI suggestion unavailable: {str(e)}"
+def get_safe_message(language):
+    safe_messages = {
+        "en": "This message appears safe.",
+        "de": "Diese Nachricht scheint sicher zu sein.",
+        "fr": "Ce message semble sûr.",
+        "es": "Este mensaje parece seguro.",
+        "it": "Questo messaggio sembra sicuro.",
+        "ta": "இந்த செய்தி பாதுகாப்பானதாக தெரிகிறது.",
+        "ja": "このメッセージは安全と思われます。",
+        "ko": "이 메시지는 안전한 것으로 보입니다.",
+        "zh-cn": "此消息似乎是安全的。",
+        "hi": "यह संदेश सुरक्षित प्रतीत होता है।"
+    }
+
+    return safe_messages.get(
+        language,
+        "This message appears safe."
+    )
 @app.route("/predict", methods=["POST"])
 def predict():
     try:
         data = request.get_json()
 
+        print("Received:", data)
+
         text = data.get("text")
-        if not text:
-            # Simple file append for warning
-            with open("api.log", "a") as f:
-                f.write(f"WARNING: No text provided at {__import__('datetime').datetime.now()}\n")
-            return jsonify({"error": "No text provided"}), 400
 
-        text_vector = vectorizer.transform([text])
+        print("Text:", text)
+        translated_text, detected_language = translate_to_english(text)
+
+        text_vector = vectorizer.transform([translated_text])
+
         prediction = model.predict(text_vector)
+
         final_output = label_encoder.inverse_transform(prediction)[0]
 
-        # Simple file append for prediction log
-        text_preview = text[:50] + "..." if len(text) > 50 else text
-        with open("api.log", "a") as f:
-            from datetime import datetime
-            f.write(f"{datetime.now()} - Prediction: '{text_preview}' -> {final_output}\n")
-        return jsonify({"input": text, "prediction": final_output})
+        final_output = label_encoder.inverse_transform(prediction)[0]
+
+        if final_output in ["spam","smishing"]:
+            suggestion = generate_ai_suggestion(
+                text,
+                final_output
+            )
+        else:
+            suggestion = get_safe_message(detected_language)
+
+        return jsonify({
+            "input": text,
+            "prediction": final_output,
+            "suggestion": suggestion
+    })
 
     except Exception as e:
-        with open("api.log", "a") as f:
-            from datetime import datetime
-            f.write(f"{datetime.now()} - ERROR: {str(e)}\n")
+        print("ERROR:", str(e))
         return jsonify({"error": str(e)}), 500
+
 
 
 if __name__ == "__main__":
     FLASK_PORT = int(os.getenv("FLASK_PORT", 5000))
-    app.run(host="0.0.0.0", port=FLASK_PORT, debug=True)
+    app.run(host="0.0.0.0", port=FLASK_PORT, debug=True)

backend/config.py

@@ -5,10 +5,6 @@
 
 DATABASE_PATH = os.getenv("DATABASE_PATH", "spam_detection.db")
 API_KEY = os.getenv("API_KEY", "")
-BASE_URL = os.getenv("BASE_URL", "http://localhost:8000")
-PORT = int(os.getenv("PORT", "5000"))
-FRONTEND_URL = os.getenv("FRONTEND_URL", "http://localhost:5173")
-DB_HOST     = os.getenv("DB_HOST",     "localhost")
-DB_USER     = os.getenv("DB_USER",     "root")
-DB_PASSWORD = os.getenv("DB_PASSWORD", "")
-DB_NAME     = os.getenv("DB_NAME",     "spam_detection")
+BASE_URL = os.getenv("BASE_URL", "http://localhost:5173")
+PORT = int(os.getenv("PORT", "8000"))
+FRONTEND_URL = os.getenv("FRONTEND_URL", "http://localhost:8000")

backend/main.py

@@ -1,107 +1,43 @@
 import os
-import joblib
-import numpy as np
-from pathlib import Path
-from fastapi import FastAPI, HTTPException
-from pydantic import BaseModel
+from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
-from backend.xai_service import XAIService
-from backend.config import FRONTEND_URL, BASE_URL, PORT
-xai_service = XAIService()
-
-# ── Resolve model paths relative to this file ────────────────────────────────
-# FIX: Use pathlib.Path so the app works regardless of the working directory.
-# Previously, hardcoded relative strings like "linear_svm_model.pkl" would
-# break whenever the process was not launched from the repo root.
-BASE_DIR = Path(__file__).resolve().parent.parent
 
-# ── Load ML models ────────────────────────────────────────────────────────────
-# FIX: label_encoder.pkl was never loaded here, causing /predict to return
-# a raw integer (0, 1, 2) instead of a human-readable label string like
-# "ham", "spam", or "smishing". The frontend's string comparisons
-# (result === "ham") would always evaluate to false with the old code.
-model         = joblib.load(BASE_DIR / "linear_svm_model.pkl")
-vectorizer    = joblib.load(BASE_DIR / "backend" / "tfidf_vectorizer.pkl")
-label_encoder = joblib.load(BASE_DIR / "label_encoder.pkl")
+# import environment config
+from backend.config import FRONTEND_URL, BASE_URL, PORT
 
 app = FastAPI(title="Spam Detection System")
 
-# ── CORS setup ────────────────────────────────────────────────────────────────
+
+# ── CORS setup (uses env variable) ─────────────────────────────
 app.add_middleware(
     CORSMiddleware,
     allow_origins=[
-        FRONTEND_URL,
-        os.getenv("FRONTEND_DEV_URL", "http://localhost:3000"),
+    FRONTEND_URL,
+    os.getenv("FRONTEND_DEV_URL", "http://localhost:8000"),
     ],
     allow_credentials=True,
     allow_methods=["*"],
     allow_headers=["*"],
 )
 
-# ── Request schema ────────────────────────────────────────────────────────────
-class PredictIn(BaseModel):
-    text: str
-    type: str
-
-# ── Prediction route ──────────────────────────────────────────────────────────
-@app.post("/predict")
-def predict(body: PredictIn):
-    """
-    Classify a message as ham, spam, or smishing.
-
-    Returns:
-        prediction (str): Human-readable label — "ham", "spam", or "smishing".
-        confidence (float): SVM decision-function score for the winning class.
-                            Higher absolute value = more confident prediction.
-    """
-    try:
-        vectorized_text = vectorizer.transform([body.text])
-
-        # Get the raw predicted class index (0, 1, or 2)
-        raw_prediction = model.predict(vectorized_text)[0]
 
-        # FIX: Convert class index → string label using the label encoder
-        label = label_encoder.inverse_transform([raw_prediction])[0]
-
-        # ENHANCEMENT: Return a confidence score.
-        # LinearSVC does not support predict_proba(); use decision_function()
-        # instead. The score for each class is its distance from the boundary —
-        # a higher value means the model is more certain of that class.
-        scores = model.decision_function(vectorized_text)[0]
-        confidence = round(float(np.max(scores)), 4)
-
-        return {
-            "prediction": label,       # e.g. "ham", "spam", "smishing"
-            "confidence": confidence,  # e.g. 1.2345
-        }
-    except Exception as exc:
-        raise HTTPException(status_code=500, detail=str(exc))
-
-# ── Health / root ─────────────────────────────────────────────────────────────
+# ── Basic health check ─────────────────────────────────────────
 @app.get("/")
 def root():
     return {
-        "status":   "ok",
-        "message":  "Spam Detection API is running",
+        "status": "ok",
+        "message": "Spam Detection API is running",
         "base_url": BASE_URL,
     }
 
+
 @app.get("/health")
 def health():
     return {"status": "healthy"}
 
-# ── Routers ───────────────────────────────────────────────────────────────────
-# EMAIL DATABASE ROUTES (Issue #13)
-from backend.emails import router as emails_router
-# from backend.database import init_db  # Uncomment once DB is configured
-# init_db()
-app.include_router(emails_router)
-
-# EXPORT ROUTES (Issue #23)
-from backend.export import router as export_router
-app.include_router(export_router)
 
-# ── Run directly ──────────────────────────────────────────────────────────────
+# ── Optional: run directly ─────────────────────────────────────
 if __name__ == "__main__":
     import uvicorn
+
     uvicorn.run("backend.main:app", host="0.0.0.0", port=PORT, reload=True)

backend/requirements.txt

@@ -2,7 +2,4 @@ Flask==3.1.0
 scikit-learn==1.6.1
 joblib==1.5.1
 python-dotenv==1.1.1
-fastapi
-uvicorn
-mysql-connector-python
-fpdf2
+python deep-translator langdetect - 1.11.4

frontend/electron.cjs

@@ -0,0 +1,17 @@
+const { BrowserWindow, app } = require("electron");
+const path = require("path");
+
+function createWindow() {
+  const win = new BrowserWindow({
+    width: 1200,
+    height: 800
+  });
+
+  const indexPath = path.join(app.getAppPath(), "dist", "index.html");
+
+  console.log("Loading:", indexPath);
+
+  win.loadFile(indexPath);
+}
+
+app.whenReady().then(createWindow);