I am trying to figure out how a higher timeout set for the openai llm option isnt clobbered by the retry_decorator here
retry = create_retry_decorator(
max_retries=max_retries,
random_exponential=True,
stop_after_delay_seconds=60,
min_seconds=1,
max_seconds=20,
)

Even though I have set a higher timeout, the retry always kicks in at 60 seconds, shouldn't the retry decorator use the same value for timeout in stop_after_delay_seconds

Maybe I am imissing something?

For additional context I am trying to call a chat completion with the deepseek r1 model, and it always returns from the POST to https://api.deepseek.com/chat/completions in 60 seconds with a 200 OK, but empty response.

Related to my previous thread on the timeout bug, it looks like it wouldnt matter even if the bug is fixed, because when they are having issues, deepseek returns in 60 seconds, with an 200 OK, but an empty response. While at the same time indicating that the response is in json. Which breaks in the openai library here, since there is no check to see if what is returned is actually json.

But since this is a deepseek issue which is an openai like api, fixing it's going to probably take a fork of the openai api, or handle it elsewhere. sigh

@Logan M Big Docs are continuing to plague me with issues. When I create a DocumentSummaryIndex, this line grabs the first node's metadata and that ends up exceeding pinecone limits. Shouldnt this also add the exclude llm/embed field lists? I did try to add that but that embed exclusion filter seems to happen somewhere else...

Hi there, have been working with ingestion pipelines, docstores and I am finding that for a large document with a large number of nodes there can be a significant performance hit when doing any document management like delete/add. This is because it does a put on every node action, in delete e.g. https://github.com/run-llama/llama_index/blob/a24292c79424affeeb47920b327c20eca5ba85ff/llama-index-core/llama_index/core/storage/docstore/keyval_docstore.py#L485), and depending on the number of remaining nodes, it can take a while. Would it make more sense to wait til all the nodes are removed before doing the put for ref_doc_info?