Platform for stateful agents: AI with advanced memory that can learn and self-improve over time.
- Add cache control to system messages - Cache static system content to reduce token costs on repeated requests - Maintain existing conversation flow while optimizing API costs - Support both string and list format system messages
Astrotalk1531829cc39d
diff --git a/letta/llm_api/anthropic_client.py b/letta/llm_api/anthropic_client.pyindex 5f790c221..cbadecaf0 100644--- a/letta/llm_api/anthropic_client.py+++ b/letta/llm_api/anthropic_client.py@@ -243,7 +243,8 @@ class AnthropicClient(LLMClientBase):# Move 'system' to the top levelif messages[0].role != "system":raise RuntimeError(f"First message is not a system message, instead has role {messages[0].role}")- data["system"] = messages[0].content if isinstance(messages[0].content, str) else messages[0].content[0].text+ system_content = messages[0].content if isinstance(messages[0].content, str) else messages[0].content[0].text+ data["system"] = self._add_cache_control_to_system_message(system_content)data["messages"] = [m.to_anthropic_dict(inner_thoughts_xml_tag=inner_thoughts_xml_tag,@@ -489,6 +490,27 @@ class AnthropicClient(LLMClientBase):)return chat_completion_response+ def _add_cache_control_to_system_message(self, system_content):+ """Add cache control to system message content"""+ if isinstance(system_content, str):+ # For string content, convert to list format with cache control+ return [+ {+ 'type': 'text',+ 'text': system_content,+ 'cache_control': {'type': 'ephemeral'}+ }+ ]+ elif isinstance(system_content, list):+ # For list content, add cache control to the last text block+ cached_content = system_content.copy()+ for i in range(len(cached_content) - 1, -1, -1):+ if cached_content[i].get('type') == 'text':+ cached_content[i]['cache_control'] = {'type': 'ephemeral'}+ break+ return cached_content++ return system_contentdef convert_tools_to_anthropic_format(tools: List[OpenAITool]) -> List[dict]: