Fixed auto approver and inference continuation

sia/web_agent.py

@@ -97,10 +97,6 @@ class WebAgent(BaseAgent):
         """Update context and reset all LLM states"""
         with self._llm_lock:
             self._context = context
-            self._response_buffer.clear()
-            for llm_name in self._llms:
-                self._set_llm_state(llm_name, LlmState.IDLE)
-            
             for handler in self._context_change_handlers:
                 handler(context, generated)
 
@@ -118,7 +114,6 @@ class WebAgent(BaseAgent):
             return self._stop_flags[llm_name]
         response_token_iter = llm.infer(self.system_prompt, self.context, self._response_buffer.get_text(), should_stop)
         for token in response_token_iter:
-            print(token, end='', flush=True)
             with self._output_lock:
                 self._response_buffer.append_text(token)
         with self._llm_lock:
@@ -132,11 +127,10 @@ class WebAgent(BaseAgent):
 
     def approve_response(self) -> None:
         """Process approved response from specified LLM"""
-        if self.llms.get(llm_name) != LlmState.IDLE:
-            return
         timestamp = datetime.now(timezone.utc)
         self._iteration_logger.log_iteration(timestamp, self._context, self._response_buffer.get_text())
         parse_result = self._parser.parse(timestamp, self._response_buffer.get_text())
+        self._response_buffer.clear()
         if isinstance(parse_result, Command):
             result = parse_result.execute(self._working_memory)
             self._command_result = result