diff --git a/app/app.py b/app/app.py
index 7b299a0..ce64a77 100644
--- a/app/app.py
+++ b/app/app.py
@@ -37,10 +37,31 @@ def notifications(ws):
 # ================
 #   DECORATORS
 # ================
+
+import hashlib
+
+def make_cache_key(model, prompt):
+    raw = f"{model}:{prompt}"
+    return hashlib.sha256(raw.encode()).hexdigest()
+
+def get_cache(model, prompt):
+    key = make_cache_key(model, prompt)
+    db = SessionLocal()
+    entry = db.query(Cache).filter_by(key=key).first()
+    db.close()
+    return entry.response if entry else None
+
+def set_cache(model, prompt, response):
+    key = make_cache_key(model, prompt)
+    db = SessionLocal()
+    entry = Cache(key=key, model=model, response=response)
+    db.add(entry)
+    db.commit()
+    db.close()
+
+
 import structlog
-
 logger = structlog.get_logger()
-
 @app.before_request
 def log_request():
     logger.info(
@@ -1013,9 +1034,65 @@ def generate():
     #     return jsonify({"error": "Invalid JSON response from Ollama."}), 502
     #output_text = ollama_data.get("response") or ollama_data.get("output") or ""
     #return jsonify({"model": model_name, "output": output_text})
-
+   
+    cached = get_cache(model, prompt)
+    if cached:
+        return jsonify({"model": model, "output": cached, "cached": True})
+    
     result, used_model = call_ollama_with_fallback({"prompt": prompt})
-    return jsonify({"model_used": used_model, "output": result.get("response")})
+    output = result.get("response")
+    
+    set_cache(used_model, prompt, output)
+    
+    return jsonify({"model": used_model, "output": output, "cached": False})
+
+@app.route("/api/multi", methods=["POST"])
+@login_required
+@rate_limited
+def multi():
+    data = request.json
+    models = data.get("models", ["llama3.2", "gemma4"])
+    prompt = data.get("prompt")
+
+    results = {}
+
+    for m in models:
+        try:
+            r = requests.post(
+                f"{app.config['OLLAMA_BASE_URL']}/api/generate",
+                json={"model": m, "prompt": prompt}
+            )
+            r.raise_for_status()
+            results[m] = r.json().get("response")
+        except Exception as e:
+            results[m] = f"Error: {str(e)}"
+
+    return jsonify(results)
+
+@app.route("/api/multi/stream")
+@login_required
+def multi_stream():
+    models = request.args.get("models", "llama3.2,gemma4").split(",")
+    prompt = request.args.get("prompt")
+
+    def generate():
+        for m in models:
+            yield f"event: model\n"
+            yield f"data: {json.dumps({'model': m})}\n\n"
+
+            url = f"{app.config['OLLAMA_BASE_URL']}/api/generate"
+            payload = {"model": m, "prompt": prompt, "stream": True}
+
+            with requests.post(url, json=payload, stream=True) as r:
+                for line in r.iter_lines():
+                    if not line:
+                        continue
+                    data = json.loads(line.decode())
+                    token = data.get("response", "")
+                    yield f"event: token\n"
+                    yield f"data: {json.dumps({'model': m, 'token': token})}\n\n"
+
+    return Response(generate(), mimetype="text/event-stream")