added embeddings

app.py

@@ -1,7 +1,10 @@
 import os
 import gradio as gr
 import redis
+import numpy as np
+import json
 from openai import AzureOpenAI
+from sentence_transformers import SentenceTransformer
 
 # Redis Cloud connection
 redis_client = redis.Redis(
@@ -9,42 +12,76 @@ redis_client = redis.Redis(
     port=12628,
     decode_responses=True,
     username="default",
-    password=os.getenv("REDIS_PASSWORD")  # store password as HF secret
+    password=os.getenv("REDIS_PASSWORD")
 )
 
-# Azure OpenAI client
+# Azure OpenAI client (only for chat, not embeddings anymore)
 client = AzureOpenAI(
     api_key=os.getenv("AZURE_OPENAI_API_KEY").strip(),
-    api_version="2025-01-01-preview",   # valid API version
+    api_version="2025-01-01-preview",
     azure_endpoint=os.getenv("AZURE_OPENAI_ENDPOINT").strip()
 )
-DEPLOYMENT_NAME = "gpt-4.1"  # must match your Azure deployment name!
+
+CHAT_DEPLOYMENT = "gpt-4.1"   # your Azure chat deployment
+
+# Load a small local HF embedding model
+embedder = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
+
+# Helper: get embedding from HF
+def get_embedding(text):
+    return embedder.encode(text, convert_to_numpy=True).astype(np.float32)
+
+# Helper: cosine similarity
+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):
+    query_vec = get_embedding(user_input)
+    best_key, best_score, best_val = None, -1, None
+
+    for key, val in redis_client.hgetall("cache").items():
+        entry = json.loads(val)
+        vec = np.array(entry["embedding"], dtype=np.float32)
+        score = cosine_similarity(query_vec, vec)
+        if score > best_score:
+            best_score, best_key, best_val = score, key, entry["output"]
+
+    if best_score >= threshold:
+        return best_val
+    return None
+
+def store_cache(user_input, output):
+    vec = get_embedding(user_input).tolist()
+    redis_client.hset("cache", user_input, json.dumps({
+        "embedding": vec,
+        "output": output
+    }))
 
 def chat_with_ai(user_input):
     if not user_input:
         return "Please type something."
 
-    # Check Redis first
-    cached = redis_client.get(user_input)
+    # 🔍 Check Redis semantic cache
+    cached = search_cache(user_input)
     if cached:
         return f"[From Redis] {cached}"
 
     # Otherwise query Azure OpenAI
     response = client.chat.completions.create(
-        model=DEPLOYMENT_NAME,
+        model=CHAT_DEPLOYMENT,
         messages=[{"role": "user", "content": user_input}],
         max_tokens=150
     )
     output = response.choices[0].message.content.strip()
 
-    # Save in Redis
-    redis_client.setex(user_input, 3600, output)
+    # 💾 Save with embedding in Redis
+    store_cache(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 Demo")
+    gr.Markdown("# 💬 Azure OpenAI + Redis Cloud (Semantic Cache) Demo")
     with gr.Row():
         chatbot = gr.Chatbot(type="messages")
     with gr.Row():
@@ -61,4 +98,4 @@ with gr.Blocks(title="Azure OpenAI + Redis Cloud Chat") as demo:
     msg.submit(respond, [msg, chatbot], [chatbot, msg])
 
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860, debug=True,pwa=True)
+    demo.launch(server_name="0.0.0.0", server_port=7860, debug=True, pwa=True)