Discover LlamaIndex: Joint Text to SQL and Semantic Search

Подписаться 20 тыс.

Просмотров 29 тыс.

50% 1

In this tutorial, we show users how to make use of a feature within LlamaIndex that allows them to join the results from a SQL database with that of a vector database, with the help of LLMs.
This combines two stacks: text-to-SQL for structured data as well as semantic search for unstructured data (both of which LlamaIndex has).

Опубликовано:

2 окт 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 29

@oaga74 3 месяца назад

What if the information is distributed in multiple tables within the SQL database? Would LlamaIndex be able to join the relevant tables without the user explicitly citing them? This question is essential, especially if the tables are not labeled more intuitively e.g. a table having cities labelled/named as Lctin!

@starmorph Год назад

excited for this whole channel!

@zeppelin2105 Год назад

hi~could u share more about the selector realization? whether i can use the llm itself to realize?

@vishnuajoshy1588 11 месяцев назад

Instead of using open AI..Can we do this using Llama or mistral model?

@vgtgoat 8 месяцев назад

Yes you can run your own server with something like Oobabooga textgen webui

@garyjo3229 6 месяцев назад

And ollama2 is posible language natural to sql ?

@deeplearning7097 11 месяцев назад

Thank you very much, very nice work. Could you please share your notebook? Thanks again.

@MrSawaiz Год назад

Can we implement the same thing in Langchain?

@vgtgoat 8 месяцев назад

Very little reason to implement it yourself with langchain when llamaindex exists here. You could recreate all this with raw langchain if you want technically though.

@MrSawaiz 8 месяцев назад

@@vgtgoat I understand that but I am having some issues while using this query engine. I asked on their discord channel and they said that they won't be fixing the bug any time soon. I thought maybe I'll implement my version

@renaudgg 8 месяцев назад

why we use pinecone exactly? i know with llama_index there is alresdy some VectorStorrIndex etc is it because pinecone does it better? i dont understand that part

@elruinnou Год назад

Is there any JS version of this? At least for the structure ones, cz I think it’d still be better if the unstructred ones can just use python

@jayasreelingam5862 6 месяцев назад

Can you share the code once

@FoxnewsFan Год назад

So it's basically some sort of an Agent

@CharlySUASISA Год назад

how can we use llamaindex on nodejs ?

@dota2gammingfun93 Год назад

How can i use it whithout OPEN API KEY - i dont want to use Open AI and want to use open source llm

@dennissdigitaldump8619 Год назад

There was a pine api key. This is Llama, which can be run locally.

@SanjayRoy-vz5ih 8 месяцев назад

@@dennissdigitaldump8619 pine is a vector database and has nothing to do with foundation model llms though...one way is to use huggingface hub api token

@DenisBougrimov Год назад

Can something similar be done with SparQL instead of SQL?

@SanjayRoy-vz5ih 8 месяцев назад

Yess, if it is pretty trained with SparkQL and has a solid understanding developed over time

@jcachat Год назад

🔥🔥🔥

@kelvinadungosi1579 5 месяцев назад

Hi what if I dont have ustructured data and instead of a vector store, I just want to query the llm itself or the general internet. Plus how to implement chat functionality where by I can ask follow up questions based on previous answers?