Are there any approaches to also extract & index Hyperlinks found in PDFs using LlamaIndex with the SimpleDirectoryReader?

Has anyone faced the problem when using the SentenceWindowNodeParser for document indexing that when setting a window_size of like 6 (default is 3), the LLM responses using that index is very long and unrelated to general questions like "Hey, how are you today?" or "Who is Elon Musk" and responds with the content & window of the top nodes?

Can someone give an updated Migration guide to v0.10.0? The migration guide linked on the website does not work: https://pretty-sodium-5e0.notion.site/v0-10-0-Migration-Guide-6ede431dcb8841b09ea171e7f133bd77

Has anyone else experienced non relevant / too broad responses when using a ChatMemoryBuffer? If the chat memory is empty, a question to content that was indexed is correctly synthesized in the LLM response. However, if I first chat a little bit about general stuff that are not in the indexed documents, then ask my question to indexed content, the LLM response is worse and not relevant to my data sources.

Is it possible to only allow responses from the LLM with information from the retrieved nodes using LlamaIndex, thus restricting the LLM from responding to general questions that are not in the data sources?

Is someone else having issues with LlamaIndex when using ElasticsearchStore with the ChatEngine (chat mode "context") and SentenceWindowNodeParser where the LLM response is sometimes incomplete and responding with cut of chunks? The mentioned issue emerged when I upgraded LlamaIndex from 0.8.47 to 0.9.4. The pipeline for my chat bot stayed the same throughout the two versions but now the responses are completely inpredictable. In my logs, I can see that it correctly retrieves the relevant nodes from my ElasticSearch index and LLM completion input seems fine. Since I'm using Azure OpenAI gpt-35-turbo, I've also adjusted the embeddings to use AzureOpenAIEmbedding instead of OpenAIEmbedding that was mentioned in the change logs. Does someone have any ideas?

Are there any viable approches for this issue: When loading all the documents for an index using a PDF loader, how can I only update the index with PDF files that have changed or have been removed using the ElasticSearchVectorstore without re-indexing all the files again, thus saving embedding costs?

Has anyone set up AzureOpenAI with LlamaIndex? I've deployed a gpt-35-turbo model and a text-embedding-ada-002 model to my Azure subscription. Also, I've initialized the AzureOpenAI & OpenAIEmbedding class with my model, apiBase, apiKey, apiType and deployment name (credentials from azure) and passed it into the service_context. Now, the chat completion goes through Azure OpenAI. However, when I look at my console logs, I see that for the embedding, the URL "api.openai.com/v1/embeddings" is called. Question: Why does it not go to Azure but the official OpenAI API?

Does anyone know if OpenAI stores the prompts, inputs, outputs etc. if it is used through their OpenAI API? Nowhere do they precisely clarify this.

Hello. Is there a way to add a static SQL statement to the NLSQLTableQueryEngine like a "LIMIT 10;"?