want to test embeddings

app.py

@@ -2,6 +2,11 @@ import gradio as gr
 from openai import OpenAI
 import os
 from dotenv import load_dotenv
+from datasets import load_dataset
+from sklearn.metrics.pairwise import cosine_similarity
+from functools import lru_cache
+import numpy as np
+
 
 load_dotenv(verbose=True)
 
@@ -10,31 +15,67 @@ client = OpenAI(
     api_key=os.getenv("OPENAI_API_KEY"),
 )
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    # Convert Gradio's history into OpenAI's format
+HF_DATASET = "carching/cs-data"
+EMBED_MODEL = "text-embedding-3-small"
+TOP_K = 5
+
+# ---- Load dataset ----
+def load_data():
+    ds = load_dataset(HF_DATASET)
+    messages = ds["train"]["message"]
+    senders = ds["train"]["sender"]
+    users = ds["train"]["user"]
+    return messages, senders, users
+
+messages, senders, users = load_data()
+
+# ---- Embedding helper ----
+@lru_cache(maxsize=None)
+def get_embedding(text):
+    resp = client.embeddings.create(
+        model=EMBED_MODEL,
+        input=text
+    )
+    return np.array(resp.data[0].embedding)
+
+# ---- Precompute dataset embeddings ----
+message_embeddings = [get_embedding(msg) for msg in messages]
+
+# ---- Retrieval ----
+def retrieve_context(user_query):
+    query_emb = get_embedding(user_query).reshape(1, -1)
+    all_embs = np.stack(message_embeddings)
+    sims = cosine_similarity(query_emb, all_embs)[0]
+    top_idx = np.argsort(sims)[-TOP_K:][::-1]
+    return [(messages[i], senders[i], sims[i]) for i in top_idx]
+
+# ---- Chatbot function ----
+def respond(message, history, system_message, max_tokens, temperature, top_p):
+    # Retrieve historical WhatsApp context
+    context_rows = retrieve_context(message)
+    context_text = "\n".join([f"{role}: {msg}" for msg, role, _ in context_rows])
+
+    # Build system prompt with context
+    full_system_message = (
+        f"{system_message}\n\n"
+        f"Use the following historical conversation context if relevant:\n\n"
+        f"{context_text}"
+    )
+
+    # Convert Gradio history into OpenAI format
+    chat_messages = [{"role": "system", "content": full_system_message}]
     for user_msg, assistant_msg in history:
         if user_msg:
-            messages.append({"role": "user", "content": user_msg})
+            chat_messages.append({"role": "user", "content": user_msg})
         if assistant_msg:
-            messages.append({"role": "assistant", "content": assistant_msg})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
+            chat_messages.append({"role": "assistant", "content": assistant_msg})
+    chat_messages.append({"role": "user", "content": message})
 
     # Stream responses from OpenAI
+    response_text = ""
     stream = client.chat.completions.create(
-        model="gpt-4o-mini",  # You can change this to another available model
-        messages=messages,
+        model="gpt-4o-mini",
+        messages=chat_messages,
         max_tokens=max_tokens,
         temperature=temperature,
         top_p=top_p,
@@ -44,15 +85,14 @@ def respond(
     for chunk in stream:
         if chunk.choices[0].delta.content:
             token = chunk.choices[0].delta.content
-            response += token
-            yield response
-
+            response_text += token
+            yield response_text
 
-# Gradio Chat Interface
+# ---- Gradio Interface ----
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+        gr.Textbox(value="You are a helpful support assistant.", label="System message"),
         gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
         gr.Slider(
@@ -60,10 +100,10 @@ demo = gr.ChatInterface(
             maximum=1.0,
             value=0.95,
             step=0.05,
-            label="Top-p (nucleus sampling)",
+            label="Top-p (nucleus sampling)"
         ),
     ],
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()

requirements.txt

@@ -1,4 +1,5 @@
 huggingface_hub
 gradio
 openai
-dotenv
+dotenv
+datasets