added new features and changed embedding model

app.py

@@ -15,8 +15,8 @@ redis_client = redis.Redis(
     password=os.getenv("REDIS_PASSWORD")
 )
 
-# 🧹 Clear Redis DB on startup
-redis_client.flushdb()
+# 🧹 Do NOT flush DB globally anymore, since multi-user support
+# redis_client.flushdb()
 
 # Azure OpenAI client (only for chat, not embeddings anymore)
 client = AzureOpenAI(
@@ -25,10 +25,10 @@ client = AzureOpenAI(
     azure_endpoint=os.getenv("AZURE_OPENAI_ENDPOINT").strip()
 )
 
-CHAT_DEPLOYMENT = "gpt-4.1"   # your Azure chat deployment
+CHAT_DEPLOYMENT = "gpt-4.1"
 
-# 🚀 Better embedding model from HF
-embedder = SentenceTransformer("sentence-transformers/all-mpnet-base-v2")
+# 🚀 Super lightweight multilingual embedding model
+embedder = SentenceTransformer("intfloat/multilingual-e5-small")
 
 # Helper: get embedding from HF
 def get_embedding(text):
@@ -38,11 +38,12 @@ def get_embedding(text):
 def cosine_similarity(vec1, vec2):
     return float(np.dot(vec1, vec2) / (np.linalg.norm(vec1) * np.linalg.norm(vec2)))
 
-def search_cache(user_input, threshold=0.8):
+def search_cache(user_id, user_input, threshold=0.8):
     query_vec = get_embedding(user_input)
     best_key, best_score, best_val = None, -1, None
+    cache_key = f"cache:{user_id}"
 
-    for key, val in redis_client.hgetall("cache").items():
+    for key, val in redis_client.hgetall(cache_key).items():
         entry = json.loads(val)
         vec = np.array(entry["embedding"], dtype=np.float32)
         score = cosine_similarity(query_vec, vec)
@@ -53,19 +54,24 @@ def search_cache(user_input, threshold=0.8):
         return best_val
     return None
 
-def store_cache(user_input, output):
+def store_cache(user_id, user_input, output):
     vec = get_embedding(user_input).tolist()
-    redis_client.hset("cache", user_input, json.dumps({
+    cache_key = f"cache:{user_id}"
+    redis_client.hset(cache_key, user_input, json.dumps({
         "embedding": vec,
         "output": output
     }))
 
-def chat_with_ai(user_input):
+def clear_user_cache(user_id):
+    cache_key = f"cache:{user_id}"
+    redis_client.delete(cache_key)
+
+def chat_with_ai(user_id, user_input):
     if not user_input:
         return "Please type something."
 
     # 🔍 Check Redis semantic cache
-    cached = search_cache(user_input)
+    cached = search_cache(user_id, user_input)
     if cached:
         return f"[From Redis] {cached}"
 
@@ -79,27 +85,37 @@ def chat_with_ai(user_input):
     output = response.choices[0].message.content.strip()
 
     # 💾 Save with embedding in Redis
-    store_cache(user_input, output)
+    store_cache(user_id, user_input, output)
 
     return f"[From OpenAI] {output}"
 
 # Gradio UI
 with gr.Blocks(title="Azure OpenAI + Redis Cloud Chat") as demo:
-    gr.Markdown("# 💬 Azure OpenAI + Redis Cloud (Semantic Cache) Demo")
+    gr.Markdown("# 💬 Azure OpenAI + Redis Cloud (Semantic Cache, Multi-User)")
+
+    user_id = gr.Textbox(label="User ID", placeholder="Enter your username", value="guest")
+
     with gr.Row():
         chatbot = gr.Chatbot(type="messages")
+
     with gr.Row():
         msg = gr.Textbox(placeholder="Type your message here...")
         send = gr.Button("Send")
+        clear = gr.Button("🧹 Clear Cache")
 
-    def respond(message, history):
-        bot_reply = chat_with_ai(message)
+    def respond(message, history, user_id):
+        bot_reply = chat_with_ai(user_id, message)
         history.append({"role": "user", "content": message})
         history.append({"role": "assistant", "content": bot_reply})
         return history, ""
 
-    send.click(respond, [msg, chatbot], [chatbot, msg])
-    msg.submit(respond, [msg, chatbot], [chatbot, msg])
+    def clear_cache_ui(user_id, history):
+        clear_user_cache(user_id)
+        return [], f"✅ Cache cleared for {user_id}"
+
+    send.click(respond, [msg, chatbot, user_id], [chatbot, msg])
+    msg.submit(respond, [msg, chatbot, user_id], [chatbot, msg])
+    clear.click(clear_cache_ui, [user_id, chatbot], [chatbot, msg])
 
 if __name__ == "__main__":
     demo.launch(server_name="0.0.0.0", server_port=7860, debug=True, pwa=True)