429 Too many requests on /api/v2/oauth with concurrent embedding invocations #275
Description
Hi.
I spin off multiple (just 2 so far) concurrent tasks for embeddings requests:
async with asyncio.TaskGroup() as tg:
for url in urls:
tg.create_task(invoke_gigachat_embeddings(url))
it seems if oauth token needs to be refreshed it's done simultaneously and falls with:
line 158, in _build_vectors
batch_embeddings = await giga_embeddings.aembed_documents(list(texts))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/app/.local/lib/python3.12/site-packages/langchain_gigachat/embeddings/gigachat.py", line 177, in aembed_documents
embeddings = await self._client.aembeddings(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/app/.local/lib/python3.12/site-packages/gigachat/client.py", line 436, in aembeddings
return await self._adecorator(_acall)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/app/.local/lib/python3.12/site-packages/gigachat/client.py", line 417, in _adecorator
await self._aupdate_token()
File "/home/app/.local/lib/python3.12/site-packages/gigachat/client.py", line 388, in _aupdate_token
self._access_token = await post_auth.asyncio(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/app/.local/lib/python3.12/site-packages/gigachat/api/post_auth.py", line 49, in asyncio
return build_response(response, AccessToken)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/app/.local/lib/python3.12/site-packages/gigachat/api/utils.py", line 88, in build_response
raise ResponseError(response.url, response.status_code, response.content, response.headers)
gigachat.exceptions.ResponseError: (URL('https://ngw.devices.sberbank.ru:9443/api/v2/oauth'), 429, b'<html>\r\n<head><title>429 Too Many Requests</title></head>\r\n<
body>\r\n<center><h1>429 Too Many Requests</h1></center>\r\n<hr><center>SynGX</center>\r\n</body>\r\n</html>\r\n', Headers({'server': 'SynGX', 'date': 'Mon, 21 Jul 2025 05:26:50 GMT', 'content-type': 'text/html', 'content-length': '162', 'connection': 'keep-alive', 'allow': 'GET, POST', 'strict-transport-security': 'max-age=3153
6000; includeSubDomains'}))
Can you suggest how to make client app code (I have one idea, btw) or the library code more resilient eg introduce retires or timeouts etc ?
Thanks