graph TD
 A[OpenClaw 發(fā)起 LLM 請求] --> B{底層 API 響應(yīng)}
 B -->|響應(yīng)時間 < timeout| C[? 正常返回結(jié)果]
 B -->|響應(yīng)時間 > timeout| D[? llm request timed out]
 D --> E[排查原因]
 E --> F[token 過長?]
 E --> G[API 端點(diǎn)慢?]
 E --> H[網(wǎng)絡(luò)問題?]
 F --> I[方案一: 調(diào)大 timeout]
 G --> J[方案二: 換低延遲端點(diǎn)]
 H --> K[方案三: Streaming + 重試]

import time
from openai import OpenAI
client = OpenAI(
 api_key="your-key",
 base_url="https://api.ofox.ai/v1" # 聚合接口，一個 Key 可調(diào)多個模型
)
start = time.time()
try:
 response = client.chat.completions.create(
 model="gpt-5",
 messages=[{"role": "user", "content": "說一句話測試連通性"}],
 max_tokens=50,
 timeout=10 # 故意設(shè)短，測試響應(yīng)速度
 )
 elapsed = time.time() - start
 print(f"? 響應(yīng)成功，耗時: {elapsed:.2f}s")
 print(f"返回內(nèi)容: {response.choices[0].message.content}")
except Exception as e:
 elapsed = time.time() - start
 print(f"? 請求失敗，耗時: {elapsed:.2f}s")
 print(f"錯誤信息: {e}")

# openclaw.yaml 或你的配置文件
llm:
 provider: openai_compatible
 model: claude-4.6-sonnet
 base_url: "https://api.ofox.ai/v1"
 api_key: "your-key"
 timeout: 300 # 單位秒，默認(rèn)60，長任務(wù)建議300
 max_retries: 3 # 超時后自動重試次數(shù)

from openclaw import Agent, LLMConfig
llm_config = LLMConfig(
 model="claude-4.6-sonnet",
 base_url="https://api.ofox.ai/v1",
 api_key="your-key",
 timeout=300,
 max_retries=3
)
agent = Agent(llm=llm_config)
result = agent.run("幫我分析這段代碼的性能瓶頸...")

from openai import OpenAI
# 換成低延遲的聚合接口
client = OpenAI(
 api_key="your-ofox-key",
 base_url="https://api.ofox.ai/v1"
)
# 其他代碼完全不用動
response = client.chat.completions.create(
 model="claude-4.6-sonnet",
 messages=[{"role": "user", "content": "你的 prompt"}],
 max_tokens=2000
)

import time
from openai import OpenAI
from tenacity import retry, stop_after_attempt, wait_exponential
client = OpenAI(
 api_key="your-key",
 base_url="https://api.ofox.ai/v1",
 timeout=300 # 總超時兜底
)
@retry(
 stop=stop_after_attempt(3),
 wait=wait_exponential(multiplier=1, min=2, max=30),
 reraise=True
)
def call_llm_with_streaming(prompt: str, model: str = "claude-4.6-sonnet") -> str:
 """帶 Streaming + 自動重試的 LLM 調(diào)用"""
 full_response = ""
 stream = client.chat.completions.create(
 model=model,
 messages=[{"role": "user", "content": prompt}],
 max_tokens=4000,
 stream=True # 關(guān)鍵！開啟流式
 )
 for chunk in stream:
 if chunk.choices[0].delta.content:
 content = chunk.choices[0].delta.content
 full_response += content
 print(content, end="", flush=True) # 實時輸出
 print() # 換行
 return full_response
# 使用
try:
 result = call_llm_with_streaming(
 "幫我寫一份完整的 API 接口設(shè)計文檔，要求覆蓋認(rèn)證、限流、錯誤碼..."
 )
 print(f"\n生成完成，總長度: {len(result)} 字符")
except Exception as e:
 print(f"重試 3 次后仍然失敗: {e}")

from openclaw import Agent, LLMConfig
llm_config = LLMConfig(
 model="claude-4.6-sonnet",
 base_url="https://api.ofox.ai/v1",
 api_key="your-key",
 timeout=300,
 max_retries=3,
 stream=True # OpenClaw 支持 stream 模式
)
agent = Agent(
 llm=llm_config,
 on_token=lambda token: print(token, end="") # 實時輸出回調(diào)
)
result = agent.run("你的復(fù)雜任務(wù) prompt...")

import httpx
client = OpenAI(
 api_key="your-key",
 base_url="https://api.ofox.ai/v1",
 http_client=httpx.Client(
 timeout=httpx.Timeout(
 connect=30.0, # 連接超時 30s
 read=300.0, # 讀取超時 300s
 write=30.0, # 寫入超時 30s
 pool=30.0 # 連接池超時 30s
 )
 )
)

import logging
logging.basicConfig(level=logging.DEBUG)
# 或者 OpenClaw 自帶的 debug 模式
agent = Agent(llm=llm_config, verbose=True)