Update app.py

app.py

@@ -1,8 +1,7 @@
 import gradio as gr
 from faster_whisper import WhisperModel
 from llama_cpp import Llama
-import requests
-import os
+from duckduckgo_search import DDGS
 import time
 
 # Initialize models
@@ -18,43 +17,26 @@ llm = Llama.from_pretrained(
     verbose=False
 )
 
-# Get Brave API key from environment
-BRAVE_API_KEY = os.getenv("BRAVE_API_KEY", "")
+# Initialize DuckDuckGo Search (no API key needed!)
+ddgs = DDGS(timeout=3)
 
 def search_web(query, max_results=3):
-    """Perform web search using Brave API"""
-    if not BRAVE_API_KEY:
-        return "⚠️ Brave API key not configured. Add it in Space Settings."
-    
+    """Perform web search using DuckDuckGo (FREE & UNLIMITED)"""
     try:
-        headers = {
-            "Accept": "application/json",
-            "Accept-Encoding": "gzip",
-            "X-Subscription-Token": BRAVE_API_KEY
-        }
-        params = {
-            "q": query,
-            "count": max_results
-        }
-        
-        response = requests.get(
-            "https://api.search.brave.com/res/v1/web/search",
-            headers=headers,
-            params=params,
-            timeout=2
+        # Use text search for fast results
+        results = ddgs.text(
+            keywords=query,
+            region='wt-wt',  # Worldwide results
+            safesearch='moderate',
+            timelimit='m',  # Last month for freshness
+            max_results=max_results
         )
         
-        if response.status_code != 200:
-            return f"Search error: {response.status_code}"
-        
-        data = response.json()
-        results = data.get("web", {}).get("results", [])
-        
         context = ""
         for i, result in enumerate(results[:max_results], 1):
-            title = result.get("title", "")
-            description = result.get("description", "")
-            context += f"\n[{i}] {title}\n{description}\n"
+            title = result.get('title', '')
+            body = result.get('body', '')
+            context += f"\n[{i}] {title}\n{body}\n"
         
         return context.strip() if context else "No search results found."
     
@@ -68,10 +50,10 @@ def process_audio(audio_path, question_text=None):
     # Step 1: Transcribe audio if provided
     if audio_path:
         try:
-            segments, _ = whisper_model.transcribe(audio_path, language="en")
+            segments, _ = whisper_model.transcribe(audio_path, language="en", beam_size=1)
             question = " ".join([seg.text for seg in segments])
         except Exception as e:
-            return f"Transcription error: {str(e)}", 0.0
+            return f"❌ Transcription error: {str(e)}", 0.0
     else:
         question = question_text
     
@@ -82,48 +64,51 @@ def process_audio(audio_path, question_text=None):
     
     # Step 2: Web search for current info
     search_start = time.time()
-    search_results = search_web(question)
+    search_results = search_web(question, max_results=2)  # Reduced to 2 for speed
     search_time = time.time() - search_start
     
     # Step 3: Generate answer with LLM
     llm_start = time.time()
-    prompt = f"""You are a helpful assistant. Answer the question briefly based on the context below.
+    prompt = f"""Answer the question briefly using the context below.
 
-Context from web search:
+Context:
 {search_results}
 
 Question: {question}
 
-Answer (be concise and accurate):"""
+Answer:"""
     
     try:
         response = llm(
             prompt,
-            max_tokens=150,
-            temperature=0.3,
-            top_p=0.9,
+            max_tokens=120,  # Reduced for faster generation
+            temperature=0.2,  # Lower for faster, more focused responses
+            top_p=0.85,
             stop=["Question:", "\n\n\n"],
             echo=False
         )
         
         answer = response['choices'][0]['text'].strip()
     except Exception as e:
-        answer = f"LLM error: {str(e)}"
+        answer = f"❌ LLM error: {str(e)}"
     
     llm_time = time.time() - llm_start
     total_time = time.time() - start_time
     
-    timing_info = f"\n\n⏱️ **Timing:** Transcription={transcription_time:.2f}s | Search={search_time:.2f}s | LLM={llm_time:.2f}s | **Total={total_time:.2f}s**"
+    # Color code timing (green if under 3s, yellow if close, red if over)
+    time_emoji = "🟢" if total_time < 3.0 else "🟡" if total_time < 3.5 else "🔴"
+    
+    timing_info = f"\n\n{time_emoji} **Timing:** Trans={transcription_time:.2f}s | Search={search_time:.2f}s | LLM={llm_time:.2f}s | **Total={total_time:.2f}s**"
     
     return answer + timing_info, total_time
 
 # Create Gradio interface
-with gr.Blocks(title="Fast Q&A with Web Search", theme=gr.themes.Soft()) as demo:
+with gr.Blocks(title="Fast Q&A - FREE Unlimited Search", theme=gr.themes.Soft()) as demo:
     gr.Markdown("""
-    # 🎤 Fast Political Q&A System
-    Ask questions via audio or text. Get web-grounded answers in ~3 seconds!
+    # ⚡ Ultra-Fast Political Q&A System
+    Ask questions via audio or text. **FREE unlimited web search** with DuckDuckGo!
     
-    **Features:** Whisper-tiny + Qwen2.5-0.5B + Brave Search API
+    **Features:** Whisper-tiny + Qwen2.5-0.5B + DuckDuckGo (No API Key!)
     """)
     
     with gr.Tab("🎙️ Audio Input"):
@@ -180,39 +165,44 @@ with gr.Blocks(title="Fast Q&A with Web Search", theme=gr.themes.Soft()) as demo
             examples=[
                 ["Who won the 2024 US presidential election?"],
                 ["What is the current inflation rate in India?"],
-                ["Who is the prime minister of UK?"]
+                ["Who is the prime minister of UK?"],
+                ["What is the latest news about AI?"]
             ],
             inputs=text_input
         )
     
-    with gr.Accordion("📡 API Usage", open=False):
+    with gr.Accordion("📡 API Usage via curl", open=False):
         gr.Markdown("""
-        ### Using curl to query this endpoint:
-        
-        **Text Query:**
+        ### Text Query (Simplest):
         ```
         curl -X POST https://archcoder-basic-app.hf.space/call/text_query \\
           -H "Content-Type: application/json" \\
           -d '{"data": ["Who is the current US president?"]}'
         ```
         
-        **Audio Query:**
+        ### Audio Query:
         ```
-        # 1. Upload audio file
+        # Upload audio
         curl -F "files=@audio.mp3" https://archcoder-basic-app.hf.space/upload
         
-        # 2. Query with returned path
+        # Query (replace path from upload response)
         curl -X POST https://archcoder-basic-app.hf.space/call/audio_query \\
           -H "Content-Type: application/json" \\
-          -d '{"data": [{"path": "/tmp/gradio/audio.mp3"}]}'
+          -d '{"data": [{"path": "/tmp/gradio/YOUR_FILE.mp3"}]}'
         ```
         """)
     
     gr.Markdown("""
     ---
-    **Note:** This Space uses free-tier resources. For production use, consider upgrading to a persistent Space.
+    ### 🎯 System Specs
+    - **Search:** DuckDuckGo (FREE, unlimited, no API key!)
+    - **Transcription:** Whisper-tiny (optimized for speed)
+    - **LLM:** Qwen2.5-0.5B Q4 (fast factual answers)
+    - **Target:** Sub-3s total response time
+    
+    🟢 = Under 3s | 🟡 = 3-3.5s | 🔴 = Over 3.5s
     """)
 
 if __name__ == "__main__":
-    demo.queue()
+    demo.queue(max_size=5)  # Limit queue for consistent performance
     demo.launch()