i figured out how to increase the max_tokens so that my answers don't get cut off. however, i don't see a way to do this if you load_from_disk your index (without creating the index in the session from scratch)....am i missing something?

Thanks for your inputs! 95% of my documents are powerpoints, so i was planning on chunking slide by slide and generating an embedding per slide. is that the same concept as using sentence transformers?

my main question though is what should the GPT-Index index structure be? because of the vast amount of data, would I need to go in a mult-level tree direction? would this hinder performance?

has anyone taken an existing elasticsearch (or any) index and retrofitted llamaindex on top of it? i am trying to do that with an existing elasticsearch index i have, but keep running into

File "C:\Python311\Lib\site-packages\llama_index\vector_stores\elasticsearch.py", line 542, in aquery
f"Could not parse metadata from hit {hit['_source']['metadata']}"
~~~~~~^^^^^^^^^^^^
KeyError: 'metadata'

understandably because the format of the metadata in my Elasticsearch index is not the expected format for the ElasticsearchStore class in LlamaIndex. how do you get around this without having to re-index everything?

hi, i am trying to build a chatbot using openai on azure. i am following these instructions: https://gpt-index.readthedocs.io/en/latest/examples/agent/openai_agent.html. i've replaced this line:



with this:



but I am getting this error:



in this line:



does anything immediately jump out? am i able to simply switch the llm to use AzureOpenAI?

does llama index support Vespa?

curious if anyone here has used gpt-index on top of a truly large corpus of data without training the model on your corpus, and whether the results were production quality? i'm in the process of trying to implement this on my corpus of data (115k docs, ~15k words per doc) using OpenAI, and I'm concerned I will need to change course and use a model I can actually train on my data because relevancy won't be good enough