Enable multiple llms

sia/base_agent.py

@@ -22,7 +22,6 @@ class BaseAgent(ABC):
             action_schema: str, 
             working_memory: WorkingMemory,
             metrics: SystemMetrics,
-            llm: LlmEngine,
             validator: XMLValidator,
             parser: ResponseParser,
         ):
@@ -33,7 +32,6 @@ class BaseAgent(ABC):
         self._action_schema = action_schema
         self._working_memory = working_memory
         self._metrics = metrics
-        self._llm = llm
         self._validator = validator
         self._parser = parser
         
@@ -47,7 +45,7 @@ class BaseAgent(ABC):
         """Get the system prompt."""
         return f"{self._system_prompt}\n{self._action_schema}"
             
-    def _compile_context(self) -> str:
+    def _compile_context(self, llmEngine: LlmEngine) -> str:
         """
         Compile the current context for LLM inference.
         Includes system metrics and working memory entries.
@@ -76,8 +74,8 @@ class BaseAgent(ABC):
         context_str = pretty_print_element(context)
         
         # Calculate token usage percentage
-        token_count = self._llm.token_count(self.system_prompt, context_str)
-        token_limit = self._llm.token_limit()
+        token_count = llmEngine.token_count(self.system_prompt, context_str)
+        token_limit = llmEngine.token_limit()
         context_usage = (float(token_count) / float(token_limit)) * 100.0
         
         # Update context usage metric