Skip to content

armgit / letta

Public Imported from GitHub
0

Platform for stateful agents: AI with advanced memory that can learn and self-improve over time.

fix: async functions must call async methods (#8612)

cthomas57cb2d7566d6
diff --git a/letta/llm_api/llm_client_base.py b/letta/llm_api/llm_client_base.py
index 00f0815e5..608451a82 100644
--- a/letta/llm_api/llm_client_base.py
+++ b/letta/llm_api/llm_client_base.py
@@ -67,7 +67,7 @@ class LLMClientBase:
try:
log_event(name="llm_request_sent", attributes=request_data)
- response_data = self.request(request_data, llm_config)
+ response_data = await self.request_async(request_data, llm_config)
if step_id and telemetry_manager:
telemetry_manager.create_provider_trace(
actor=self.actor,
diff --git a/letta/schemas/memory.py b/letta/schemas/memory.py
index 982a97aef..4a9e64e4e 100644
--- a/letta/schemas/memory.py
+++ b/letta/schemas/memory.py
@@ -318,14 +318,6 @@ class Memory(BaseModel, validate_assignment=True):
llm_config=llm_config,
)
- @trace_method
- async def compile_in_thread_async(self, tool_usage_rules=None, sources=None, max_files_open=None, llm_config=None) -> str:
- """Deprecated: use compile() instead."""
- import warnings
-
- logger.warning("compile_in_thread_async is deprecated; use compile()", stacklevel=2)
- return self.compile(tool_usage_rules=tool_usage_rules, sources=sources, max_files_open=max_files_open, llm_config=llm_config)
-
def list_block_labels(self) -> List[str]:
"""Return a list of the block names held inside the memory object"""
return [block.label for block in self.blocks]
diff --git a/letta/services/passage_manager.py b/letta/services/passage_manager.py
index cbd305496..99781b555 100644
--- a/letta/services/passage_manager.py
+++ b/letta/services/passage_manager.py
@@ -1,9 +1,8 @@
import uuid
from datetime import datetime, timezone
-from functools import lru_cache
from typing import Dict, List, Optional
-from openai import AsyncOpenAI, OpenAI
+from openai import AsyncOpenAI
from sqlalchemy import func, select
from sqlalchemy.ext.asyncio import AsyncSession
from sqlalchemy.orm import noload
@@ -31,16 +30,6 @@ from letta.utils import enforce_types
logger = get_logger(__name__)
-# TODO: Add redis-backed caching for backend
-@lru_cache(maxsize=8192)
-def get_openai_embedding(text: str, model: str, endpoint: str) -> List[float]:
- from letta.settings import model_settings
-
- client = OpenAI(api_key=model_settings.openai_api_key, base_url=endpoint, max_retries=0)
- response = client.embeddings.create(input=text, model=model)
- return response.data[0].embedding
-
-
@async_redis_cache(key_func=lambda text, model, endpoint: f"{model}:{endpoint}:{text}")
async def get_openai_embedding_async(text: str, model: str, endpoint: str) -> list[float]:
from letta.settings import model_settings