Hi, I am new to llama index and trying

hhimanshu0113

Hi, I am new to llama index and trying to build a basic document retrieval system. I am using Azure OpenAI embeddings. I have two problems and I am not sure how to resolve:

I have a long documents with multiple paragraphs. I want to to treat each paragraph as a seperate document. How to do this?

I have many text files. I can create documents object for each. But how do I generate embeddings? Everytime with more one 1 document, I get error as “Too many inputs…..” Seems like some limitation of Azure embeddings. How to resolve this?

Thanks in advance!!

23 comments

WWhiteFang_Jr

Hi,

If you want to create document for each paragraph separately you can either use Textnode class or a wrapper over the Textnode class in the form of Document

Plain Text

from llama_index.schema import TextNode

para_1 = TextNode(text="Your text here")

# To insert this into a index
index = VectorStoreIndex([para_1])

How are you generating embeddings here?

hhimanshu0113

I am using AzureOpenAIEmbedding. Now whenever I am calling VectorStoreIndex with multiple documents, it is giving me error.

WWhiteFang_Jr

Can you share your code for the same, if possible?

hhimanshu0113

Plain Text

from llama_index.embeddings import AzureOpenAIEmbedding

embed_model = AzureOpenAIEmbedding(
    model="text-embedding-ada-002",
    deployment_name="textembeddingada002",
    api_key=api_key,
    azure_endpoint=azure_endpoint,
    api_version=api_version,
    
)

from llama_index import set_global_service_context

service_context = ServiceContext.from_defaults(
    llm=llm,
    embed_model=embed_model,
)

set_global_service_context(service_context)

documents = SimpleDirectoryReader(
    input_files=["./data/paul_graham_essay.txt"]
).load_data()

index = VectorStoreIndex.from_documents(documents)

hhimanshu0113

On the last line, I get

" Too many inputs. The max number of inputs is 1. We hope to increase the number of inputs per request soon...."

WWhiteFang_Jr

Code looks fine, Can you check what do you get in the documents variable

hhimanshu0113

Plain Text

[Document(id_='84d29648-c597-407c-968a-443924ebf956', embedding=None, metadata={'file_path': 'data/paul_graham_essay.txt', 'file_name': 'paul_graham_essay.txt', 'file_type': 'text/plain', 'file_size': 75042, 'creation_date': '2023-12-11', 'last_modified_date': '2023-12-11', 'last_accessed_date': '2023-12-11'}, excluded_embed_metadata_keys=['file_name', 'file_type', 'file_size', 'creation_date', 'last_modified_date', 'last_accessed_date'], excluded_llm_metadata_keys=['file_name', 'file_type', 'file_size', 'creation_date', 'last_modified_date', 'last_accessed_date'], relationships={}, hash='5dfe27179663d7ae4c02bbb134d50a62143a55545a90f194a20454deb5df5901', text='\n\nWhat I Worked On\n\nFebruary 2021\n\nBefore college the two main things I worked on, outside of school, ...................Jessica Livingston, Robert Morris, and Harj Taggar for reading drafts of this.\n', start_char_idx=None, end_char_idx=None, text_template='{metadata_str}\n\n{content}', metadata_template='{key}: {value}', metadata_seperator='\n')]

hhimanshu0113

I omited the text here. It is text sample from llama website only. Multiple paragraphs.

hhimanshu0113

It is single Document element here. ^
When I just keep a single paragraph here, it works fine.

hhimanshu0113

Do I have to define chunk_size or similar value somewhere?

WWhiteFang_Jr

No default is fine, Can you share your full error ? as the code looks alright to me

Also what is the version that you are trying with

LLogan M

This is classic azure, set embed_batch_size=1 in the embedding model

hhimanshu0113

Oh, is that so. I will try out this, Thanks!
Anyway, is there any documentation available on AzureOpenAIEmbedding class?

LLogan M

Not really documentation, but an example!

LLogan M

https://docs.llamaindex.ai/en/stable/examples/customization/llms/AzureOpenAI.html

hhimanshu0113

Yes I saw this. But looking for documentation with all the parameters and explanations.

LLogan M

Best bet is api docs I guess

https://docs.llamaindex.ai/en/stable/api_reference/llms/azure_openai.html

LLogan M

But tbh api docs suck, reading source code is more informative lol

LLogan M

Attachment

hhimanshu0113

True, thanx!

hhimanshu0113

Is it the case that when a document/text is large enough, the AzureOpenAIEmbedding object will break the text into multiple batches and create embeddings then?
Context: While using embed_batch_size=1 with a small document length, it works fine. When I put in a large text document, thenembedding generation fails again with "Too many requests" error.

hhimanshu0113

Creating nodes from the documents is right approch in this case?

LLogan M

Right, from_documents() will chunk documents into nodes, or you can do it manually

Add a reply

Find answers from the community

Hi, I am new to llama index and trying