设置了AI网关的大模型地址后，启动openclaw实例，里面没有大模型地址

[5309609]

设置了AI网关的大模型地址后，启动openclaw实例，里面没有大模型地址，应该是带进去才对吧？

[Iamlovingit]

@5309609 can you provide the instance's env? Like this:

env |grep CLAWMANAGER_LLM_BASE_URL
CLAWMANAGER_LLM_BASE_URL=http://clawmanager-gateway.clawmanager-system.svc.cluster.local:9001/api/v1/gateway/llm

[5309609]

@5309609 can you provide the instance's env? Like this:@5309609 你能提供该实例的环境信息吗？格式如下：

env |grep CLAWMANAGER_LLM_BASE_URL
CLAWMANAGER_LLM_BASE_URL=http://clawmanager-gateway.clawmanager-system.svc.cluster.local:9001/api/v1/gateway/llm

[5309609]

openclaw的pod日志一直在打印
2026/04/08 11:47:20 load config: bootstrap_token is required
2026/04/08 11:47:21 load config: bootstrap_token is required
2026/04/08 11:47:22 load config: bootstrap_token is required
2026/04/08 11:47:23 load config: bootstrap_token is required
2026/04/08 11:47:24 load config: bootstrap_token is required

[Iamlovingit]

openclaw的pod日志一直在打印
2026/04/08 11:47:20 load config: bootstrap_token is required
2026/04/08 11:47:21 load config: bootstrap_token is required
2026/04/08 11:47:22 load config: bootstrap_token is required
2026/04/08 11:47:23 load config: bootstrap_token is required
2026/04/08 11:47:24 load config: bootstrap_token is required

@5309609 This looks like env missing CLAWMANAGER_AGENT_BOOTSTRAP_TOKEN
Please show your env：
env |grep CLAWMANAGER_AGENT_BOOTSTRAP_TOKEN

[5309609]

openclaw的pod日志一直在打印
2026/04/08 11:47:20 load config: bootstrap_token is required
2026/04/08 11:47:21 load config: bootstrap_token is required
2026/04/08 11:47:22 load config: bootstrap_token is required
2026/04/08 11:47:23 load config: bootstrap_token is required
2026/04/08 11:47:24 load config: bootstrap_token is required

@5309609 This looks like env missing CLAWMANAGER_AGENT_BOOTSTRAP_TOKEN Please show your env： env |grep CLAWMANAGER_AGENT_BOOTSTRAP_TOKEN

升级到4.9的版本后，报错改变了，env |grep CLAWMANAGER_AGENT_BOOTSTRAP_TOKEN
CLAWMANAGER_AGENT_BOOTSTRAP_TOKEN=agt_boot_ca2cbc343d6742a7863bedd0c0ce5a858ba79ef5bff4684f
是存在的，openclaw中报错LLM request timed out.
pod日志中无报错日志，大模型的接口地址是公司内网使用vllm部署的qwen3.5-122b

[Iamlovingit]

@5309609 Sorry for late reply.
The most likely causes are:

• the ClawManager pod cannot reach the internal vLLM endpoint from inside the cluster
• the upstream qwen3.5-122b response is exceeding the current AI Gateway timeout

For the next round of checking, please provide:

• the configured vLLM base URL
• whether that URL is reachable from inside the ClawManager pod
• the corresponding AI Gateway invocation / failure record for the failed request