Add oscar gradio

app.py

@@ -1,21 +1,21 @@
 import os
+import re
 import tempfile
 
 import gradio as gr
 import numpy as np
 import soundfile as sf
 import torch
+from ddgs import DDGS
 from huggingface_hub import hf_hub_download
+from llama_cpp import Llama
 from transformers import pipeline
 from TTS.api import TTS
 
-from llama_cpp import Llama
-
-
 # --- Device Setup ---
-device = "cpu" 
+device = "cpu"
 
-# --- 1. STT Setup (Whisper) --- 
+# --- 1. STT Setup (Whisper) ---
 print("Loading Whisper...")
 STT_MODEL_NAME = "openai/whisper-tiny.en"
 stt_pipe = pipeline("automatic-speech-recognition", model=STT_MODEL_NAME, device=device)
@@ -47,13 +47,49 @@ tts_model = TTS(model_name=TTS_MODEL_NAME, progress_bar=False)
 
 
 # --- Core Functions ---
+def get_web_context(message):
+    search_keywords = [
+        "current",
+        "latest",
+        "recent",
+        "today",
+        "now",
+        "news",
+        "weather",
+        "price",
+        "2024",
+        "2025",
+        "what is happening",
+        "score",
+        "match",
+    ]
+
+    if not any(keyword in message.lower() for keyword in search_keywords):
+        return None
+
+    try:
+        with DDGS() as ddgs:
+            results = list(ddgs.text(message, max_results=3))
+
+        if not results:
+            print("No search results found")
+            return None
+
+        print(f"Found {len(results)} results:")
+        context = "Current information from web search:\n"
+        for i, result in enumerate(results):
+            print(f"Result {i+1}: {result['title']}")
+            print(f"  Body: {result['body'][:100]}...")
+            context += f"- {result['title']}: {result['body'][:200]}...\n"
+
+        return context
+
+    except Exception as e:
+        print(f"Search error: {e}")
+        return None
 
 
 def chat_with_bot(message, history):
-    """
-    Chat with your model using HuggingFace InferenceClient.
-    """
-    # Ensure history is a list
     if history is None:
         history = []
 
@@ -61,32 +97,67 @@ def chat_with_bot(message, history):
         return history, ""
 
     try:
+        web_context = get_web_context(message=message)
+
         # Build conversation context from history
-        context = ""
+        conversation = ""
         for h in history:
             role = "User" if h.get("role") == "user" else "Assistant"
-            context += f"{role}: {h.get('content', '')}\n"
-        
-        # Create prompt with context
-        prompt = context + f"User: {message}\nAssistant:"
+            conversation += f"{role}: {h.get('content', '')}\n"
+
+        # Create a clearer prompt with system instruction
+        if web_context:
+            prompt = f"""Answer ONLY using this information:
+
+                {web_context}
+
+                Question: {message}
+                Answer:"""
+            print("The web context has been added to the prompt")
+        else:
+            prompt = f"""You are a helpful assistant. Answer naturally and conversationally.
+                    {conversation}User: {message}
+                    Assistant:"""
 
         print(f"Generating response with Llama...")
-        
-        # Generate response using llama.cpp
+
+        # Generate response with stricter settings
         response = llm(
             prompt,
-            max_tokens=256,
+            max_tokens=200,
             temperature=0.7,
             top_p=0.95,
+            stop=["User:", "\nUser:"],
         )
-        
+
         response_str = response["choices"][0]["text"].strip()
-        
+
+        response_str = response_str.strip("'\"")
+        response_str = response_str.rstrip(",:;")
+        response_str = response_str.strip("'\"")
+        response_str = re.sub(r"(\d+\.){10,}", "", response_str)
+
+        if "User:" in response_str:
+            response_str = response_str.split("User:")[0].strip()
+
+        response_str = response_str.replace("[{", "").replace("}]", "")
+        response_str = response_str.replace("'text':", "").replace('"text":', "")
+        response_str = response_str.replace("'type': 'text'", "").replace(
+            '"type": "text"', ""
+        )
+
+        if ", 'type'" in response_str or ', "type"' in response_str:
+            response_str = (
+                response_str.split(", 'type'")[0].split(', "type"')[0].strip()
+            )
+
+        # One final strip
+        response_str = response_str.strip("'\",:;")
+
         if not response_str:
             response_str = "I received an empty response. Please try again."
             print("Warning: Empty response from LLM")
 
-        # Append to history
         history.append({"role": "user", "content": message})
         history.append({"role": "assistant", "content": response_str})
 
@@ -94,12 +165,13 @@ def chat_with_bot(message, history):
 
     except Exception as e:
         import traceback
+
         error_trace = traceback.format_exc()
         print(f"LLM Error: {e}")
         print(f"Full traceback:\n{error_trace}")
-        
+
         error_msg = f"Error generating response: {str(e) if str(e) else 'Unknown error occurred'}"
-        
+
         history.append({"role": "user", "content": message})
         history.append({"role": "assistant", "content": error_msg})
         return history, error_msg
@@ -241,4 +313,4 @@ with gr.Blocks() as demo:
             )
             clear_btn.click(lambda: [], None, chatbot)
 
-demo.launch(css=custom_css)
+demo.launch()