Platform for stateful agents: AI with advanced memory that can learn and self-improve over time.
feat: add llm config to ttfs logging (#1998)
cthomas043283f8aa23
diff --git a/letta/server/rest_api/utils.py b/letta/server/rest_api/utils.pyindex d8f962a8b..ab20557ef 100644--- a/letta/server/rest_api/utils.py+++ b/letta/server/rest_api/utils.py@@ -20,6 +20,7 @@ from letta.helpers.message_helper import convert_message_creates_to_messagesfrom letta.log import get_loggerfrom letta.schemas.enums import MessageRolefrom letta.schemas.letta_message_content import OmittedReasoningContent, ReasoningContent, RedactedReasoningContent, TextContent+from letta.schemas.llm_config import LLMConfigfrom letta.schemas.message import Message, MessageCreatefrom letta.schemas.usage import LettaUsageStatisticsfrom letta.schemas.user import User@@ -53,6 +54,7 @@ async def sse_async_generator(usage_task: Optional[asyncio.Task] = None,finish_message=True,request_start_timestamp_ns: Optional[int] = None,+ llm_config: Optional[LLMConfig] = None,):"""Wraps a generator for use in Server-Sent Events (SSE), handling errors and ensuring a completion message.@@ -70,6 +72,7 @@ async def sse_async_generator(ttft_span = Noneif request_start_timestamp_ns is not None:ttft_span = tracer.start_span("time_to_first_token", start_time=request_start_timestamp_ns)+ ttft_span.set_attributes({f"llm_config.{k}": v for k, v in llm_config.model_dump().items()})try:async for chunk in generator:diff --git a/letta/server/server.py b/letta/server/server.pyindex 4117adf38..e8ef2d079 100644--- a/letta/server/server.py+++ b/letta/server/server.py@@ -1725,6 +1725,7 @@ class SyncServer(Server):usage_task=task,finish_message=include_final_message,request_start_timestamp_ns=request_start_timestamp_ns,+ llm_config=llm_config,),media_type="text/event-stream",)