New web interface, move llm engine to separate process

sia/base_agent.py

@@ -4,7 +4,6 @@ import xml.etree.ElementTree as ET
 from .llm_engine import LlmEngine
 from .response_parser import ResponseParser
 from .system_metrics import SystemMetrics
-from .util import pretty_print_element
 from .working_memory import WorkingMemory
 
 class BaseAgent(ABC):
@@ -36,8 +35,13 @@ class BaseAgent(ABC):
     def system_prompt(self) -> str:
         """Get the system prompt."""
         return f"{self._system_prompt}\n{self._action_schema}"
+    
+    @property
+    def working_memory(self) -> WorkingMemory:
+        """Get the working memory."""
+        return self._working_memory
             
-    def _compile_context(self, llmEngine: LlmEngine) -> str:
+    def _compile_context(self, llmEngine: LlmEngine) -> ET:
         """
         Compile the current context for LLM inference.
         Includes system metrics and working memory entries.
@@ -60,15 +64,13 @@ class BaseAgent(ABC):
 
         for entry in memory_context:
             context.append(entry)
-
-        context_str = pretty_print_element(context)
         
         # Calculate token usage percentage
-        token_count = llmEngine.token_count(self.system_prompt, context_str)
+        token_count = llmEngine.token_count(self.system_prompt, context)
         token_limit = llmEngine.token_limit()
         context_usage = (float(token_count) / float(token_limit)) * 100.0
         
         # Update context usage metric
         context.set("context", f"{str(round(context_usage, 2))}%")
         
-        return pretty_print_element(context)
+        return context