Update app.py

app.py

@@ -1,13 +1,15 @@
 import gradio as gr
 from faster_whisper import WhisperModel
 from llama_cpp import Llama
-from brave import Brave
+import requests
 import os
 import time
 
 # Initialize models
-print("Loading models...")
+print("Loading Whisper model...")
 whisper_model = WhisperModel("tiny", device="cpu", compute_type="int8")
+
+print("Loading LLM...")
 llm = Llama.from_pretrained(
     repo_id="Qwen/Qwen2.5-0.5B-Instruct-GGUF",
     filename="qwen2.5-0.5b-instruct-q4_k_m.gguf",
@@ -16,19 +18,46 @@ llm = Llama.from_pretrained(
     verbose=False
 )
 
-# Initialize Brave Search
-brave_client = Brave(api_key=os.getenv("BRAVE_API_KEY", ""))
+# Get Brave API key from environment
+BRAVE_API_KEY = os.getenv("BRAVE_API_KEY", "")
 
 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."
+    
     try:
-        results = brave_client.search(q=query, count=max_results)
-        web_results = results.web_results if hasattr(results, 'web_results') else []
+        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
+        )
+        
+        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(web_results[:max_results], 1):
-            context += f"\n[{i}] {result.title}\n{result.description}\n"
-        return context.strip()
+        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"
+        
+        return context.strip() if context else "No search results found."
+    
     except Exception as e:
         return f"Search failed: {str(e)}"
 
@@ -38,59 +67,82 @@ def process_audio(audio_path, question_text=None):
     
     # Step 1: Transcribe audio if provided
     if audio_path:
-        segments, _ = whisper_model.transcribe(audio_path, language="en")
-        question = " ".join([seg.text for seg in segments])
+        try:
+            segments, _ = whisper_model.transcribe(audio_path, language="en")
+            question = " ".join([seg.text for seg in segments])
+        except Exception as e:
+            return f"Transcription error: {str(e)}", 0.0
     else:
         question = question_text
     
-    if not question:
-        return "No input provided", 0.0
+    if not question or question.strip() == "":
+        return "❌ No input provided", 0.0
     
     transcription_time = time.time() - start_time
     
-    # Step 2: Web search for political/current info
+    # Step 2: Web search for current info
     search_start = time.time()
     search_results = search_web(question)
     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 based on the context below.
+    prompt = f"""You are a helpful assistant. Answer the question briefly based on the context below.
 
 Context from web search:
 {search_results}
 
 Question: {question}
 
-Answer briefly and accurately:"""
-    
-    response = llm(
-        prompt,
-        max_tokens=150,
-        temperature=0.3,
-        top_p=0.9,
-        stop=["Question:", "\n\n"],
-        echo=False
-    )
-    
-    answer = response['choices'][0]['text'].strip()
-    llm_time = time.time() - llm_start
+Answer (be concise and accurate):"""
     
+    try:
+        response = llm(
+            prompt,
+            max_tokens=150,
+            temperature=0.3,
+            top_p=0.9,
+            stop=["Question:", "\n\n\n"],
+            echo=False
+        )
+        
+        answer = response['choices'][0]['text'].strip()
+    except Exception as 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"
+    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**"
     
     return answer + timing_info, total_time
 
 # Create Gradio interface
-with gr.Blocks(title="Fast Q&A with Web Search") as demo:
-    gr.Markdown("# 🎤 Fast Political Q&A System\nAsk questions via audio or text. Answers in ~3 seconds!")
-    
-    with gr.Tab("Audio Input"):
-        audio_input = gr.Audio(type="filepath", label="Record or upload audio question")
-        audio_submit = gr.Button("Submit Audio", variant="primary")
-        audio_output = gr.Textbox(label="Answer", lines=6)
-        audio_time = gr.Number(label="Response Time (seconds)")
+with gr.Blocks(title="Fast Q&A with Web 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!
+    
+    **Features:** Whisper-tiny + Qwen2.5-0.5B + Brave Search API
+    """)
+    
+    with gr.Tab("🎙️ Audio Input"):
+        with gr.Row():
+            with gr.Column():
+                audio_input = gr.Audio(
+                    sources=["microphone", "upload"],
+                    type="filepath",
+                    label="Record or upload audio"
+                )
+                audio_submit = gr.Button("🚀 Submit Audio", variant="primary", size="lg")
+            
+            with gr.Column():
+                audio_output = gr.Textbox(
+                    label="Answer",
+                    lines=8,
+                    show_copy_button=True
+                )
+                audio_time = gr.Number(label="Response Time (seconds)", precision=2)
         
         audio_submit.click(
             fn=lambda x: process_audio(x, None),
@@ -99,11 +151,23 @@ with gr.Blocks(title="Fast Q&A with Web Search") as demo:
             api_name="audio_query"
         )
     
-    with gr.Tab("Text Input"):
-        text_input = gr.Textbox(label="Type your question", placeholder="Who won the 2024 elections?")
-        text_submit = gr.Button("Submit Text", variant="primary")
-        text_output = gr.Textbox(label="Answer", lines=6)
-        text_time = gr.Number(label="Response Time (seconds)")
+    with gr.Tab("✍️ Text Input"):
+        with gr.Row():
+            with gr.Column():
+                text_input = gr.Textbox(
+                    label="Type your question",
+                    placeholder="Who is the current US president?",
+                    lines=3
+                )
+                text_submit = gr.Button("🚀 Submit Text", variant="primary", size="lg")
+            
+            with gr.Column():
+                text_output = gr.Textbox(
+                    label="Answer",
+                    lines=8,
+                    show_copy_button=True
+                )
+                text_time = gr.Number(label="Response Time (seconds)", precision=2)
         
         text_submit.click(
             fn=lambda x: process_audio(None, x),
@@ -111,19 +175,44 @@ with gr.Blocks(title="Fast Q&A with Web Search") as demo:
             outputs=[text_output, text_time],
             api_name="text_query"
         )
+        
+        gr.Examples(
+            examples=[
+                ["Who won the 2024 US presidential election?"],
+                ["What is the current inflation rate in India?"],
+                ["Who is the prime minister of UK?"]
+            ],
+            inputs=text_input
+        )
+    
+    with gr.Accordion("📡 API Usage", open=False):
+        gr.Markdown("""
+        ### Using curl to query this endpoint:
+        
+        **Text Query:**
+        ```
+        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:**
+        ```
+        # 1. Upload audio file
+        curl -F "files=@audio.mp3" https://archcoder-basic-app.hf.space/upload
+        
+        # 2. Query with returned path
+        curl -X POST https://archcoder-basic-app.hf.space/call/audio_query \\
+          -H "Content-Type: application/json" \\
+          -d '{"data": [{"path": "/tmp/gradio/audio.mp3"}]}'
+        ```
+        """)
     
     gr.Markdown("""
-    ### 📡 API Usage
-    ```
-    # Upload audio file
-    curl -F "files=@audio.mp3" https://YOUR-SPACE-URL/upload
-    
-    # Make query
-    curl -X POST https://YOUR-SPACE-URL/call/audio_query \\
-      -H "Content-Type: application/json" \\
-      -d '{"data": [{"path": "/tmp/uploaded_audio.mp3"}]}'
-    ```
+    ---
+    **Note:** This Space uses free-tier resources. For production use, consider upgrading to a persistent Space.
     """)
 
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860)
+    demo.queue()
+    demo.launch()