Update app.py

app.py

@@ -16,10 +16,9 @@ def get_text_from_url(url):
     soup = BeautifulSoup(response.text, 'html.parser')
     texts = soup.find_all(text=True)
     visible_texts = filter(tag_visible, texts)
-    # Only join non-empty lines
-    return "\n".join(t.strip() for t in visible_texts if t.strip())
+    return "\n".join(t.strip() for t in visible_texts)
 
-# Collect homepage text from various URL extensions
+# Get the text from your homepage (and any additional extensions as needed)
 text_list = []
 homepage_url = "https://sites.google.com/view/abhilashnandy/home/"
 extensions = ["", "pmrf-profile-page"]
@@ -27,51 +26,53 @@ for ext in extensions:
     url_text = get_text_from_url(homepage_url + ext)
     text_list.append(url_text)
 
-# Build the system message containing homepage info
+# Optionally, repeat for sub-links if necessary
+
+# Build a system message with the homepage info
 SYSTEM_MESSAGE = (
     "You are a QA chatbot to answer queries (in less than 30 words) on my homepage that has the following information -\n\n"
     + "\n\n".join(text_list)
     + "\n\n"
 )
 
-# Use a lightweight model for low-latency CPU inference
-client = InferenceClient("microsoft/DialoGPT-small")
+# Use a model that works well on CPU, has a decently long context, and low inference latency.
+# Here we choose a small chat-optimized model:
+client = InferenceClient("TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF")
 
 def respond(message, history: list[tuple[str, str]], system_message=SYSTEM_MESSAGE,
             max_tokens=140, temperature=0.7, top_p=0.95):
-    # Construct the prompt including the system message and conversation history.
-    prompt = system_message
-    for user_q, bot_a in history:
-        prompt += f"Question: {user_q}\n"
-        prompt += f"Answer: {bot_a}\n"
-    prompt += f"Question: {message}\nAnswer:"
-
+    messages = [{"role": "system", "content": system_message}]
+    for val in history:
+        if len(val) >= 1:
+            messages.append({"role": "user", "content": "Question: " + val[0]})
+        if len(val) >= 2:
+            messages.append({"role": "assistant", "content": "Answer: " + val[1]})
+    messages.append({"role": "user", "content": message})
     try:
-        response = client.text_generation(
-            prompt,
-            max_new_tokens=max_tokens,
+        response = client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
             temperature=temperature,
             top_p=top_p,
+            # stream=True,  # Uncomment to enable streaming
         )
-        # Expecting a list of dict(s) with key "generated_text"
-        generated_text = response[0]["generated_text"]
-        # Attempt to extract the answer by splitting at "Answer:"
-        answer = generated_text.split("Answer:")[-1].strip().split("\n")[0].strip()
-        return answer
+        return response.choices[0].message["content"]
     except Exception as e:
         print(f"An error occurred: {e}")
         return str(e)
 
+initial_message = [("user", "Yo who dis Abhilash?")]
 markdown_note = "## Ask Anything About Me! (Might show a tad bit of hallucination!)"
 
 demo = gr.Blocks()
-
 with demo:
     gr.Markdown(markdown_note)
     gr.ChatInterface(
-        respond,
+        fn=respond,
         examples=["Yo who dis Abhilash?", "What is Abhilash's most recent publication?"],
-        additional_inputs=[],
+        additional_inputs=[
+            # You can add extra Gradio components here if needed.
+        ],
     )
 
 if __name__ == "__main__":