Fimeg/letta-server
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

fix: add explicit timeouts to httpx clients to prevent ReadTimeout errors (#8538)

Browse Source 
This commit addresses the httpx.ReadTimeout error detected in production
by adding explicit timeout configurations to several httpx client usages:

1. MCP SSE client: Pass mcp_connect_to_server_timeout (30s) to sse_client()
2. MCP StreamableHTTP client: Pass mcp_connect_to_server_timeout (30s) to streamablehttp_client()
3. OpenAI model list API: Add 30s timeout with 10s connect timeout
4. Google AI model list/details API: Add 30s timeout with 10s connect timeout

Previously, these httpx clients were created without explicit timeouts,
which could cause ReadTimeout errors when remote servers are slow to respond.

Fixes #8073

🤖 Generated with [Letta Code](https://letta.com)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: datadog-official[bot] <datadog-official[bot]@users.noreply.github.com>
Co-authored-by: Kian Jones <11655409+kianjones9@users.noreply.github.com>
This commit is contained in:
github-actions[bot]
2026-01-13 18:22:04 -08:00
committed by Sarah Wooders
parent 85c40c8154
commit 5fbf8f93e2
4 changed files with 18 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
letta/llm_api/google_ai_client.py
View File

@@ -82,7 +82,8 @@ async def google_ai_get_model_list_async(
# Determine if we need to close the client at the end
close_client = False
if client is None:
client = httpx.AsyncClient()
# Use explicit timeout to prevent httpx.ReadTimeout errors
client = httpx.AsyncClient(timeout=httpx.Timeout(30.0, connect=10.0))
close_client = True
try:
@@ -129,7 +130,8 @@ async def google_ai_get_model_details_async(
# Determine if we need to close the client at the end
close_client = False
if client is None:
client = httpx.AsyncClient()
# Use explicit timeout to prevent httpx.ReadTimeout errors
client = httpx.AsyncClient(timeout=httpx.Timeout(30.0, connect=10.0))
close_client = True
try:

3
letta/llm_api/openai.py
View File

@@ -72,7 +72,8 @@ async def openai_get_model_list_async(
# Use provided client or create a new one
close_client = False
if client is None:
client = httpx.AsyncClient()
# Use explicit timeout to prevent httpx.ReadTimeout errors
client = httpx.AsyncClient(timeout=httpx.Timeout(30.0, connect=10.0))
close_client = True
try:

7
letta/services/mcp/sse_client.py
View File

@@ -7,6 +7,7 @@ from mcp.client.sse import sse_client
from letta.functions.mcp_client.types import SSEServerConfig
from letta.log import get_logger
from letta.services.mcp.base_client import AsyncBaseMCPClient
from letta.settings import tool_settings
# see: https://modelcontextprotocol.io/quickstart/user
MCP_CONFIG_TOPLEVEL_KEY = "mcpServers"
@@ -33,10 +34,12 @@ class AsyncSSEMCPClient(AsyncBaseMCPClient):
headers[self.AGENT_ID_HEADER] = self.agent_id
# Use OAuth provider if available, otherwise use regular headers
# Pass timeout to prevent httpx.ReadTimeout errors on slow connections
timeout = tool_settings.mcp_connect_to_server_timeout
if self.oauth_provider:
sse_cm = sse_client(url=server_config.server_url, headers=headers if headers else None, auth=self.oauth_provider)
sse_cm = sse_client(url=server_config.server_url, headers=headers if headers else None, auth=self.oauth_provider, timeout=timeout)
else:
sse_cm = sse_client(url=server_config.server_url, headers=headers if headers else None)
sse_cm = sse_client(url=server_config.server_url, headers=headers if headers else None, timeout=timeout)
sse_transport = await self.exit_stack.enter_async_context(sse_cm)
self.stdio, self.write = sse_transport

10
letta/services/mcp/streamable_http_client.py
View File

@@ -1,3 +1,4 @@
from datetime import timedelta
from typing import Optional
from mcp import ClientSession
@@ -7,6 +8,7 @@ from mcp.client.streamable_http import streamablehttp_client
from letta.functions.mcp_client.types import BaseServerConfig, StreamableHTTPServerConfig
from letta.log import get_logger
from letta.services.mcp.base_client import AsyncBaseMCPClient
from letta.settings import tool_settings
logger = get_logger(__name__)
@@ -38,16 +40,18 @@ class AsyncStreamableHTTPMCPClient(AsyncBaseMCPClient):
headers[self.AGENT_ID_HEADER] = self.agent_id
# Use OAuth provider if available, otherwise use regular headers
# Pass timeout to prevent httpx.ReadTimeout errors on slow connections
timeout = timedelta(seconds=tool_settings.mcp_connect_to_server_timeout)
if self.oauth_provider:
streamable_http_cm = streamablehttp_client(
server_config.server_url, headers=headers if headers else None, auth=self.oauth_provider
server_config.server_url, headers=headers if headers else None, auth=self.oauth_provider, timeout=timeout
)
else:
# Use streamablehttp_client context manager with headers if provided
if headers:
streamable_http_cm = streamablehttp_client(server_config.server_url, headers=headers)
streamable_http_cm = streamablehttp_client(server_config.server_url, headers=headers, timeout=timeout)
else:
streamable_http_cm = streamablehttp_client(server_config.server_url)
streamable_http_cm = streamablehttp_client(server_config.server_url, timeout=timeout)
read_stream, write_stream, _ = await self.exit_stack.enter_async_context(streamable_http_cm)