Feishu Channel LLM Request Timeout: Web UI Works But Feishu Times Out
Regression in OpenClaw v2026.3.9 where Feishu channel embedded agents timeout on LLM requests while web UI chat functions normally, indicating a channel-specific timeout or streaming configuration issue.
π Symptoms
Primary Error Manifestation
The Feishu channel consistently reports the following error during conversation:
runId=xxxxxxxxx isError=true error=LLM request timed outEnvironment Context
- Install Method: Docker container on macOS
- Backend: vLLM running via Docker networking
- Routing Chain:
openclaw -> docker -> vllm
Behavioral Differential
| Channel | LLM Request Status | Behavior |
|---|---|---|
| Web UI Chat | β Success | Normal streaming response |
| Feishu Embed | β Timeout | Request hangs then fails |
Observed Log Pattern
The timeout occurs specifically during the embedded run agent execution phase, not during initial handshake or authentication. The Feishu bot receives the request, initiates the run agent, but the LLM backend never responds within the expected timeframe.
Diagnostic Evidence
Screenshots from the issue indicate:
- Error overlay displaying
LLM request timed outin the Feishu chat interface - Blue-themed code/config view showing the internal agent configuration
- The error propagates from the embedded run agent to the Feishu message handler
π§ Root Cause
Technical Analysis
The timeout occurs due to a pipeline divergence between the Web UI and Feishu channel execution paths. Specifically:
1. Stream Mode Incompatibility with vLLM
The Feishu channel handler invokes the embedded run agent with streaming enabled by default. When the vLLM backend receives streaming requests through the Docker internal network, connection idle timeout triggers before the first token arrives.