feat: Add live streaming to Chat page

- Replace blocking execute_agent_task() with live subprocess streaming
- Use Popen() to read opencode output line-by-line in real-time
- Send 'chunk' events to frontend as agent thinks
- Frontend appends chunks incrementally for live response
- Matches Orchestrator's streaming UX
- No more waiting for complete response before seeing output

app.py

@@ -2198,7 +2198,7 @@ def chat():
 
 @app.route('/chat/send', methods=['POST'])
 def chat_send():
-    """Führt einen Agent aus und gibt die Antwort per Server-Sent Events zurück."""
+    """Führt einen Agent aus und gibt die Antwort per Server-Sent Events LIVE zurück."""
     data = request.get_json()
     prompt = data.get('prompt', '').strip()
     agent_key = data.get('agent', '').strip()
@@ -2216,17 +2216,69 @@ def chat_send():
     def generate():
         # Agent-Info senden
         yield f"data: {json.dumps({'type': 'agent_selected', 'agent': agent_name, 'agent_key': agent_key})}\n\n"
-        yield f"data: {json.dumps({'type': 'processing', 'message': f'⏳ {agent_name} arbeitet...'})}\n\n"
+        yield f"data: {json.dumps({'type': 'processing', 'message': f'⏳ {agent_name} denkt nach...'})}\n\n"
         
         try:
-            # Agent ausführen (mit Memory und Work-Dir)
-            response = execute_agent_task(agent_key, prompt)
+            # Agent live ausführen mit Streaming
+            response_text = ""
             
-            # Antwort streamen
-            yield f"data: {json.dumps({'type': 'response', 'text': response})}\n\n"
+            # System-Prompt vorbereiten
+            system_prompt = get_agent_prompt(agent_key)
+            if not system_prompt:
+                yield f"data: {json.dumps({'type': 'error', 'message': f'Kein System-Prompt für Agent {agent_key}'})}\n\n"
+                return
+            
+            dirs = ensure_agent_structure(agent_key)
+            work_dir = dirs['work_dir']
+            memory_summary = get_agent_memory_summary(agent_key)
+            kb_file = os.path.join(os.path.dirname(__file__), 'diversityball_knowledge.md')
+            
+            team_summary = ""
+            if agent_key == 'orchestrator':
+                team_summary = "\n\n" + get_team_member_summary()
+            
+            full_system = f"""{system_prompt}
+
+## Deine Erinnerungen:
+{memory_summary}{team_summary}
+
+## Wissensdatenbank:
+Die Wissensdatenbank liegt unter: {kb_file}
+- Du hast Zugriff auf das Internet via WebFetch-Tool
+- Dein Arbeitsverzeichnis: {work_dir}"""
+            
+            combined_message = f"{full_system}\n\n---\n\n{prompt}"
+            model = get_agent_model(agent_key)
+            
+            # OpenCode mit Streaming aufrufen
+            process = subprocess.Popen(
+                ['opencode', 'run', '--model', model, '--format', 'json', combined_message],
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE,
+                text=True,
+                cwd=work_dir
+            )
+            
+            # Live-Output lesen und streamen
+            for line in process.stdout:
+                try:
+                    data_json = json.loads(line.strip())
+                    if data_json.get('part', {}).get('type') == 'text':
+                        chunk = data_json.get('part', {}).get('text', '')
+                        response_text += chunk
+                        # Chunk live an Frontend senden
+                        yield f"data: {json.dumps({'type': 'chunk', 'text': chunk})}\n\n"
+                except (json.JSONDecodeError, KeyError):
+                    pass
+            
+            process.wait()
+            
+            # Agent-Kommandos parsen
+            if response_text:
+                parse_agent_commands(agent_key, response_text)
             
             # Erfolg melden
-            yield f"data: {json.dumps({'type': 'complete', 'message': '✓ Fertig', 'timestamp': datetime.now().strftime('%Y-%m-%d %H:%M'), 'response': response})}\n\n"
+            yield f"data: {json.dumps({'type': 'complete', 'message': '✓ Fertig', 'timestamp': datetime.now().strftime('%Y-%m-%d %H:%M'), 'response': response_text})}\n\n"
         
         except Exception as e:
             logger.error(f"[Chat] Fehler beim Ausführen von {agent_key}: {str(e)}")