34534

app.py

@@ -43,23 +43,27 @@ class ResearchOrchestrator:
             
             logging.info(f"Retrieved {len(search_results)} results")
             
-            # Step 3: Analyze content
-            progress(0.5, desc="🧠 Analyzing search results...")
-            yield "🧠 Analyzing search results...\n\n⏳ The AI model may be initializing. This could take a few minutes if it's the first request..."
-            analysis = self.analyzer.analyze(query, search_results)
+            # Step 3: Analyze content with streaming
+            progress(0.5, desc="🧠 Analyzing search results (streaming)...")
+            yield "## 🧠 AI Analysis (Live Streaming)\n\n"
+            
+            # Collect all streamed content
+            full_analysis = ""
+            for chunk in self.analyzer.analyze_stream(query, search_results):
+                full_analysis += chunk
+                yield chunk
             
             # Check if analysis was successful
-            if analysis.startswith("⚠️") or analysis.startswith("Analysis failed"):
-                logging.warning(f"Analysis failed: {analysis}")
+            if full_analysis.startswith("⚠️") or full_analysis.startswith("Analysis failed"):
+                logging.warning(f"Analysis failed: {full_analysis}")
                 progress(0.8, desc="⚠️ Analysis failed")
-                yield analysis
                 return
             
-            logging.info("Analysis completed successfully")
+            logging.info("Analysis streaming completed successfully")
             
             # Step 4: Manage citations
             progress(0.8, desc="📎 Adding citations...")
-            cited_analysis = self.citation_manager.add_citations(analysis, search_results)
+            cited_analysis = self.citation_manager.add_citations(full_analysis, search_results)
             logging.info("Citations added")
             
             # Step 5: Format output
@@ -119,7 +123,7 @@ def initialize_modules():
 # Initialize orchestrator
 orchestrator = initialize_modules()
 
-# Custom CSS for spinner
+# Custom CSS for spinner and streaming
 custom_css = """
 .spinner {
   border: 4px solid #f3f3f3;
@@ -131,6 +135,14 @@ custom_css = """
   display: inline-block;
   margin-right: 8px;
 }
+.streaming-content {
+  white-space: pre-wrap;
+  font-family: monospace;
+  background-color: #f8f9fa;
+  padding: 10px;
+  border-radius: 5px;
+  margin: 10px 0;
+}
 @keyframes spin {
   0% { transform: rotate(0deg); }
   100% { transform: rotate(360deg); }
@@ -158,7 +170,7 @@ with gr.Blocks(css=custom_css, title="Research Assistant") as demo:
             submit_btn = gr.Button("Research", variant="primary")
             
         with gr.Column():
-            output = gr.Markdown(label="Analysis Results")
+            output = gr.Markdown(label="Analysis Results", elem_classes=["streaming-content"])
     
     # Status indicator with spinner
     status_indicator = gr.HTML("<div id='status'><span class='spinner'></span> Ready for your research query</div>")

modules/analyzer.py

@@ -80,9 +80,80 @@ class Analyzer:
         logging.warning("⏰ Server initialization timeout reached")
         return False
 
+    def analyze_stream(self, query, search_results):
+        """
+        Analyze search results using the custom LLM with streaming output
+        Yields chunks of the response as they arrive
+        """
+        # Prepare context from search results
+        context = "\n\n".join([
+            f"Source: {result.get('url', 'N/A')}\nTitle: {result.get('title', 'N/A')}\nContent: {result.get('content', 'N/A')}"
+            for result in search_results[:5]  # Limit to top 5 for context
+        ])
+        
+        prompt = f"""
+        You are an expert research analyst. Analyze the following query and information to provide a comprehensive summary.
+
+        Query: {query}
+
+        Information:
+        {context}
+
+        Please provide:
+        1. A brief overview of the topic
+        2. Key findings or developments
+        3. Different perspectives or approaches
+        4. Potential implications or future directions
+        5. Any controversies or conflicting viewpoints
+
+        Structure your response clearly with these sections. If there is insufficient information, state that clearly.
+        """
+        
+        try:
+            # First check if server is ready
+            logging.info("Checking if server is ready for analysis...")
+            if not self.wait_for_server(timeout=180):  # 3 minutes timeout
+                error_msg = "⚠️ The AI model server is still initializing. Please try again in a few minutes."
+                logging.warning(error_msg)
+                yield error_msg
+                return
+            
+            logging.info("Server is ready. Sending streaming request to AI model...")
+            
+            # Send streaming request
+            response = self.client.chat.completions.create(
+                model="DavidAU/OpenAi-GPT-oss-20b-abliterated-uncensored-NEO-Imatrix-gguf",
+                messages=[
+                    {"role": "system", "content": "You are a helpful research assistant that provides structured, analytical responses."},
+                    {"role": "user", "content": prompt}
+                ],
+                temperature=0.7,
+                max_tokens=1500,
+                stream=True  # Enable streaming
+            )
+            
+            # Yield chunks as they arrive
+            for chunk in response:
+                if chunk.choices[0].delta.content:
+                    yield chunk.choices[0].delta.content
+                    
+            logging.info("Analysis streaming completed successfully")
+            
+        except Exception as e:
+            error_msg = str(e)
+            logging.error(f"Analysis streaming failed: {error_msg}")
+            if "503" in error_msg or "Service Unavailable" in error_msg:
+                yield "⚠️ The AI model server is currently unavailable. It may be initializing. Please try again in a few minutes."
+            elif "timeout" in error_msg.lower() or "read timeout" in error_msg.lower():
+                yield "⚠️ The AI model request timed out. The server may be overloaded or initializing. Please try again in a few minutes."
+            elif "404" in error_msg:
+                yield "⚠️ The AI model endpoint was not found. Please check the configuration."
+            else:
+                yield f"Analysis failed: {str(e)}"
+
     def analyze(self, query, search_results):
         """
-        Analyze search results using the custom LLM
+        Non-streaming version for compatibility
         """
         # Prepare context from search results
         context = "\n\n".join([