client.py:530 - LLM call timed out after 120.0s (attempt 2/3)

[againcui]

请问采用nanobot测试delegate-task技能一直报模型超时，具体如下：
nanobot is thinking...09:09:02 - LiteLLM:WARNING: get_model_cost_map.py:271 - LiteLLM: Failed to fetch remote model cost map from https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json: [Errno 101] Network is unreachable. Falling back to local backup.
2026-04-03 09:09:02.278 [WARNING ] get_model_cost_map.py:271 - LiteLLM: Failed to fetch remote model cost map from https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json: [Errno 101] Network is unreachable. Falling back to local backup.
⠋ nanobot is thinking...2026-04-03 09:09:03.830 [INFO ] nanobot.py:157 - Auto-detected LLM credentials from nanobot config (/root/.nanobot/config.json), provider matched for model='dashscope/qwen3.5-plus'
2026-04-03 09:09:03.831 [INFO ] resolver.py:143 - LLM kwargs resolved (source=OPENSPACE_LLM_* env): {'api_key': '..', 'api_base': 'https://dashscope.aliyuncs.com/compatible-mode/v1'}
2026-04-03 09:09:03.831 [INFO ] tool_layer.py:106 - Initializing OpenSpace...
2026-04-03 09:09:03.831 [INFO ] tool_layer.py:117 - ✓ LLM Client: dashscope/qwen3.5-plus
2026-04-03 09:09:03.831 [INFO ] loader.py:47 - Loaded configuration file:
2026-04-03 09:09:03.831 [INFO ] loader.py:47 - Loaded configuration file:
2026-04-03 09:09:03.832 [INFO ] loader.py:47 - Loaded configuration file:
2026-04-03 09:09:03.832 [INFO ] tool_layer.py:161 - Loaded GroundingAgent config from config_agents.json (max_iterations=30, visual_analysis_timeout=60.0s)
2026-04-03 09:09:03.832 [INFO ] tool_layer.py:181 - Skipping backends not in scope: ['web', 'gui']
2026-04-03 09:09:03.834 [INFO ] provider.py:36 - Shell security policy: allow_shell_commands=True, blocked_commands=['rm', '-rf', 'shutdown', 'reboot', 'poweroff', 'halt', 'mkfs', 'mkfs.ext4', 'mkfs.xfs', 'dd', 'iptables', 'ip6tables', 'nftables', 'systemctl', 'service', 'fdisk', 'parted', 'gdisk', 'mount', 'umount', 'chmod', '777', 'chown', 'root', 'passwd', 'useradd', 'userdel', 'usermod', 'kill', '-9', 'pkill', 'killall']
⠙ nanobot is thinking...2026-04-03 09:09:03.875 [INFO ] store.py:161 - Loaded 0 quality records from SQLite (global_count=0)
2026-04-03 09:09:03.875 [INFO ] manager.py:73 - ToolQualityManager initialized (persistence=True, records=0, global_count=0, evolve_interval=5)
2026-04-03 09:09:03.875 [INFO ] grounding_client.py:149 - ToolQualityManager initialized (records=0)
2026-04-03 09:09:03.880 [INFO ] provider.py:65 - Shell backend using LOCAL mode (no server required)
2026-04-03 09:09:03.887 [INFO ] provider.py:107 - MCPProvider initialized with 0 servers (eager=False)
2026-04-03 09:09:03.887 [INFO ] tool_layer.py:191 - ✓ Grounding Client: 3 backends
2026-04-03 09:09:03.887 [INFO ] grounding_client.py:220 - GroundingClient: RecordingManager updated
2026-04-03 09:09:03.887 [INFO ] tool_layer.py:208 - ✓ Recording enabled: 1 backends
2026-04-03 09:09:03.887 [INFO ] base.py:45 - Initialized GroundingAgent: OpenSpace-GroundingAgent
2026-04-03 09:09:03.887 [INFO ] grounding_agent.py:75 - Grounding Agent initialized: OpenSpace-GroundingAgent
2026-04-03 09:09:03.887 [INFO ] grounding_agent.py:76 - Backend scope: ['shell', 'mcp', 'system']
2026-04-03 09:09:03.887 [INFO ] grounding_agent.py:77 - Max iterations: 30
2026-04-03 09:09:03.887 [INFO ] grounding_agent.py:78 - Visual analysis timeout: 60.0s
2026-04-03 09:09:03.887 [INFO ] tool_layer.py:235 - ✓ GroundingAgent: shell, mcp, system
2026-04-03 09:09:03.887 [INFO ] tool_layer.py:622 - Host skill dir (from env): /nanobot-main/nanobot/skills
2026-04-03 09:09:03.896 [INFO ] registry.py:173 - Skill discovery complete: 10 skill(s) from 2 dir(s)
2026-04-03 09:09:03.896 [INFO ] tool_layer.py:242 - ✓ Skills: 10 discovered
2026-04-03 09:09:03.896 [INFO ] grounding_agent.py:120 - Skill registry attached (10 skill(s) available for mid-iteration retrieval)
2026-04-03 09:09:03.903 [INFO ] tool_layer.py:271 - ✓ Execution analysis enabled
2026-04-03 09:09:03.903 [INFO ] tool_layer.py:286 - ✓ Skill evolution enabled (concurrent=3)
2026-04-03 09:09:03.903 [INFO ] tool_layer.py:294 - ============================================================
2026-04-03 09:09:03.903 [INFO ] tool_layer.py:295 - OpenSpace ready to use!
2026-04-03 09:09:03.903 [INFO ] tool_layer.py:296 - ============================================================
2026-04-03 09:09:03.903 [INFO ] mcp_server.py:175 - OpenSpace engine ready (model=dashscope/qwen3.5-plus).
2026-04-03 09:09:03.912 [INFO ] mcp_server.py:183 - Auto-registered host skill dirs from OPENSPACE_HOST_SKILL_DIRS:
2026-04-03 09:09:03.920 [INFO ] tool_layer.py:345 - ============================================================
2026-04-03 09:09:03.920 [INFO ] tool_layer.py:346 - Task: 1+1 等于几？请直接计算并回答。...
2026-04-03 09:09:03.920 [INFO ] tool_layer.py:347 - ============================================================
2026-04-03 09:09:03.920 [INFO ] tool_layer.py:356 - Task ID: task_83fe4481e12f
2026-04-03 09:09:03.932 [INFO ] manager.py:533 - Recording started:
2026-04-03 09:09:03.934 [INFO ] tool_layer.py:377 - Recording started: task_83fe4481e12f
2026-04-03 09:09:03.934 [INFO ] tool_layer.py:384 - Workspace: /dataagent_skills/OpenSpace
⠹ nanobot is thinking...09:09:03 - LiteLLM:INFO: utils.py:3995 -
LiteLLM completion() model= qwen3.5-plus; provider = dashscope
2026-04-03 09:09:03.946 [INFO ] utils.py:3995 -
LiteLLM completion() model= qwen3.5-plus; provider = dashscope
⠇ nanobot is thinking...2026-04-03 09:09:30.057 [ERROR ] client.py:530 - LLM call timed out after 120.0s (attempt 2/3)
2026-04-03 09:09:30.058 [INFO ] client.py:536 - Retrying after 1.0s delay...
⠋ nanobot is thinking...09:09:31 - LiteLLM:INFO: utils.py:3995 -
LiteLLM completion() model= qwen3.5-plus; provider = dashscope
2026-04-03 09:09:31.060 [INFO ] utils.py:3995 -
LiteLLM completion() model= qwen3.5-plus; provider = dashscope
⠹ nanobot is thinking...2026-04-03 09:11:03.947 [ERROR ] client.py:530 - LLM call timed out after 120.0s (attempt 1/3)
2026-04-03 09:11:03.947 [INFO ] client.py:536 - Retrying after 1.0s delay...
⠼ nanobot is thinking...09:11:04 - LiteLLM:INFO: utils.py:3995 -
LiteLLM completion() model= qwen3.5-plus; provider = dashscope
2026-04-03 09:11:04.950 [INFO ] utils.py:3995 -
LiteLLM completion() model= qwen3.5-plus; provider = dashscope
⠋ nanobot is thinking...2026-04-03 09:11:31.061 [ERROR ] client.py:530 - LLM call timed out after 120.0s (attempt 3/3)
2026-04-03 09:11:31.061 [ERROR ] analyzer.py:766 - Analysis LLM call failed (iter 0): LLM call timed out after 120.0s

[tluma]

根据文档，可以通过 OPENSPACE_LLM_CONFIG 环境变量配置 litellm 的超时时间。

解决方案：设置更长的 request_timeout 参数。

在启动 nanobot 前添加环境变量：

export OPENSPACE_LLM_CONFIG='{"request_timeout": 300}'

或者在你的 .env 文件中添加：

OPENSPACE_LLM_CONFIG={"request_timeout": 300}

request_timeout 的单位是秒。上例设置为 300 秒（5分钟）。根据你的网络情况调整这个值。

注意：当前日志显示的是 120s 超时，这看起来是默认的 litellm timeout。如果模型响应确实很慢，增加这个值应该能解决问题。

[hailmary-ship-it]

Looking at this issue, this could be a ai api side timeout issue. Maybe try npx ai-doctor. It can fix missing timeout guards with safe request timeouts plus timeout and retry protection.