Update app.py

app.py

@@ -11,21 +11,20 @@ from llama_cpp import Llama
 import requests
 from pathlib import Path
 
-# ---------------------- Downloading model ----------------------
+# ---------------------- Download Model ----------------------
 
 MODEL_URL = "https://huggingface.co/datasets/psy7743/llama3-8b-instruct-Q8_0.gguf/resolve/main/llama3-8b-instruct-Q8_0.gguf"
 MODEL_PATH = "llama3-8b-instruct-Q8_0.gguf"
 
 if not Path(MODEL_PATH).exists():
-    print("📥 Downloading LLaMA model from Hugging Face Datasets...")
+    print("📥 Downloading LLaMA model...")
     response = requests.get(MODEL_URL, stream=True)
     with open(MODEL_PATH, "wb") as f:
         for chunk in response.iter_content(chunk_size=8192):
-            if chunk:
-                f.write(chunk)
+            f.write(chunk)
     print("✅ Download complete!")
 
-# ---------------------- Data Loading & Cleaning ----------------------
+# ---------------------- Load Data ----------------------
 
 df = pd.read_csv("jupiter_faqs.csv")
 
@@ -39,25 +38,23 @@ df['clean_question'] = df['question'].apply(clean_text)
 df['clean_answer'] = df['answer'].apply(clean_text)
 df['document'] = df.apply(lambda row: f"Question: {row['clean_question']}\nAnswer: {row['clean_answer']}", axis=1)
 
-# ---------------------- Embedding Model ----------------------
+# ---------------------- Embeddings ----------------------
 
 embedding_model = SentenceTransformer('all-mpnet-base-v2')
 df['embedding'] = df['clean_question'].apply(lambda x: embedding_model.encode(x).tolist())
 df['uid'] = [str(uuid.uuid4()) for _ in range(len(df))]
 
-# ---------------------- Persistent ChromaDB Setup ----------------------
+# ---------------------- ChromaDB ----------------------
 
 persist_dir = "chroma_qa_db"
 chroma_client = chromadb.PersistentClient(path=persist_dir, settings=Settings())
 collection_name = "qa_collection"
 
-# Reset if exists
 if collection_name in [c.name for c in chroma_client.list_collections()]:
     chroma_client.delete_collection(name=collection_name)
 
 collection = chroma_client.get_or_create_collection(name=collection_name)
 
-# Add data to collection if empty
 if len(collection.get()["ids"]) == 0:
     collection.add(
         documents=df['document'].tolist(),
@@ -65,7 +62,7 @@ if len(collection.get()["ids"]) == 0:
         ids=df['uid'].astype(str).tolist()
     )
 
-# ---------------------- LLaMA Model Load ----------------------
+# ---------------------- LLaMA ----------------------
 
 llm = Llama(
     model_path=MODEL_PATH,
@@ -74,7 +71,7 @@ llm = Llama(
     n_gpu_layers=-1,
 )
 
-# ---------------------- Retrieval & Inference ----------------------
+# ---------------------- Inference ----------------------
 
 def search_chroma(query, n_results=5):
     query_embedding = embedding_model.encode(query).tolist()
@@ -85,7 +82,7 @@ def search_chroma(query, n_results=5):
     )
     return results["documents"][0]
 
-def get_inference_system(user_query):
+def generate_response(user_query: str) -> str:
     docs = search_chroma(user_query)
     context_str = "\n\n".join(docs)
 
@@ -97,13 +94,7 @@ def get_inference_system(user_query):
     - Do not hallucinate or make up answers.
     - Keep the tone friendly."""
 
-    prompt = (
-        sys_prompt
-        + "\n\n"
-        + "context:\n" + context_str
-        + "\n\nQuestion: " + user_query
-        + "\nAnswer:"
-    )
+    prompt = f"{sys_prompt}\n\ncontext:\n{context_str}\n\nQuestion: {user_query}\nAnswer:"
 
     response = llm(
         prompt,
@@ -113,39 +104,26 @@ def get_inference_system(user_query):
         stop=["Q:", "\n"],
         echo=True
     )
-
     return response["choices"][0]["text"].split("Answer:")[-1].strip()
 
 # ---------------------- Gradio Interface ----------------------
 
-all_suggestions = list(df['clean_question'])
-
-def autocomplete_suggestions(prefix):
-    prefix = prefix.lower()
-    return [s for s in all_suggestions if s.startswith(prefix)][:5]
-
-def llama_chat(user_input, history):
-    response = get_inference_system(user_input)
-    history.append({"role": "user", "content": user_input})
-    history.append({"role": "assistant", "content": response})
-    return history, gr.update(value="")
-
-with gr.Blocks() as demo:
-    gr.Markdown("## 🦙 LLaMA-3 FAQ Bot — Ask me anything about Jupiter Money!")
-
-    chatbot = gr.Chatbot(label="LLaMA Chat", type='messages')
-    state = gr.State([])
-
-    with gr.Column():
-        txt = gr.Textbox(show_label=False, placeholder="Type your question...")
-        suggestion_md = gr.Markdown("")  # Use Markdown instead of HighlightedText
-
-    def update_suggestions(text):
-        suggestions = autocomplete_suggestions(text)
-        suggestion_text = "**Suggestions:**\n" + "\n".join(f"- {s}" for s in suggestions) if suggestions else ""
-        return gr.Markdown.update(value=suggestion_text)
-
-    txt.change(fn=update_suggestions, inputs=txt, outputs=suggestion_md)
-    txt.submit(fn=llama_chat, inputs=[txt, state], outputs=[chatbot, txt])
+def gradio_chat_interface(message, history):
+    reply = generate_response(message)
+    history = history + [(message, reply)]
+    return history, history
+
+demo = gr.ChatInterface(
+    fn=generate_response,
+    title="🦙 LLaMA-3 FAQ Chatbot",
+    chatbot=gr.Chatbot(label="Ask me anything about Jupiter Money!"),
+    examples=[
+        "What is Jupiter Edge credit card?",
+        "What happens if I miss a payment?",
+        "How to change billing address?"
+    ],
+    cache_examples=False
+)
 
-demo.launch(share=True)
+if __name__ == "__main__":
+    demo.launch(share=True)