在大多数 LLM 应用中，您会希望流式传输输出，以减少用户看到第一个 token 的时间。

LangSmith 的追踪功能原生支持通过 generator 函数进行流式输出。下面是一个示例。

from langsmith import traceable
@traceable
def my_generator():
  for chunk in ["Hello", "World", "!"]:
      yield chunk
# Stream to the user
for output in my_generator():
  print(output)
# It also works with async functions

@traceable
async def my_async_generator():
  for chunk in ["Hello", "World", "!"]:
      yield chunk
# Stream to the user
async def main():
  async for output in my_async_generator():
      print(output)
asyncio.run(main())

const myGenerator = traceable(function* () {
  for (const chunk of ["Hello", "World", "!"]) {
      yield chunk;
  }
});
for (const output of myGenerator()) {
  console.log(output);
}

聚合结果

默认情况下， outputs 会被聚合到 LangSmith 中的单个数组中。如果您想自定义其存储方式（例如，将输出连接成单个字符串），可以使用 aggregate 选项（reduce_fn 在 python 中）。这对于聚合流式 LLM 输出特别有用。

from langsmith import traceable
def concatenate_strings(outputs: list):
  return "".join(outputs)
@traceable(reduce_fn=concatenate_strings)
def my_generator():
  for chunk in ["Hello", "World", "!"]:
      yield chunk
# Stream to the user
for output in my_generator():
  print(output)
# It also works with async functions

@traceable(reduce_fn=concatenate_strings)
async def my_async_generator():
  for chunk in ["Hello", "World", "!"]:
      yield chunk
# Stream to the user
async def main():
  async for output in my_async_generator():
      print(output)
asyncio.run(main())

const concatenateStrings = (outputs: string[]) => outputs.join("");
const myGenerator = traceable(function* () {
  for (const chunk of ["Hello", "World", "!"]) {
      yield chunk;
  }
}, { aggregator: concatenateStrings });
for (const output of await myGenerator()) {
  console.log(output);
}