定制超时与速率限制

在使用 collections API 与 llm_factory 时，直接在你的 LLM client 上配置超时和重试。

OpenAI Client 配置

from openai import AsyncOpenAI
from ragas.llms import llm_factory
from ragas.metrics.collections import Faithfulness

# Configure timeout and retries on the client
client = AsyncOpenAI(
    timeout=60.0,        # 60 second timeout
    max_retries=5,       # Retry up to 5 times on failures
)

llm = llm_factory("gpt-4o-mini", client=client)

# Use with metrics
scorer = Faithfulness(llm=llm)
result = scorer.score(
    user_input="When was the first super bowl?",
    response="The first superbowl was held on Jan 15, 1967",
    retrieved_contexts=[
        "The First AFL–NFL World Championship Game was an American football game played on January 15, 1967, at the Los Angeles Memorial Coliseum in Los Angeles."
    ]
)

可用选项

细粒度超时控制

要对不同类型的超时有更多控制：

import httpx
from openai import AsyncOpenAI

client = AsyncOpenAI(
    timeout=httpx.Timeout(
        60.0,           # Total timeout
        connect=5.0,    # Connection timeout
        read=30.0,      # Read timeout
        write=10.0,     # Write timeout
    ),
    max_retries=3,
)

提供商文档

每个 LLM 提供商都有自己的 client 配置选项。请参考你的提供商的 SDK 文档：

• OpenAI Python SDK
• Anthropic Python SDK

旧版 Metrics API

以下示例使用带 RunConfig 的旧版 metrics API 模式。对于新项目，我们建议使用上文所示的基于 collections 的 API，在 client 级别进行配置。

弃用时间线

此 API 将在 0.4 版本中弃用，并在 1.0 版本中移除。请迁移到基于 collections 的 API。

RunConfig 参数

from ragas.run_config import RunConfig

run_config = RunConfig(
    timeout=180,        # Max seconds per operation (default: 180)
    max_retries=10,     # Retry attempts (default: 10)
    max_wait=60,        # Max seconds between retries (default: 60)
    max_workers=16,     # Concurrent workers (default: 16)
    log_tenacity=False, # Log retry attempts (default: False)
    seed=42,            # Random seed (default: 42)
)

与 Evaluate 一起使用

from langchain_openai import ChatOpenAI
from ragas.llms import LangchainLLMWrapper
from ragas import EvaluationDataset, SingleTurnSample, evaluate
from ragas.metrics import Faithfulness
from ragas.run_config import RunConfig

# Legacy LLM setup
llm = LangchainLLMWrapper(ChatOpenAI(model="gpt-4o"))

# Configure run settings
run_config = RunConfig(max_workers=64, timeout=60)

# Use with evaluate
results = evaluate(
    dataset=eval_dataset,
    metrics=[Faithfulness(llm=llm)],
    run_config=run_config,
)