[Feature Request]: RAG emmbeddings & 挂外部文档 about chatgpt-next-web HOT 5 OPEN

Bot detected the issue body's language is not English, translate it automatically.

Title: [Feature Request]: RAG emmbeddings & linked external documents

from chatgpt-next-web.

Problem Description

when writing some personal document I need my own doc, if there are RAG & Emmbedding function to let me upload my personal file it will be very helpful. 当写自己私有文档时需要自己的其他文档作为参考，如果可以提供挂载外部文档的功能将非常有用。

Solution Description

add vector db like chromadb to save local files. 增加向量数据库比如chromadb用来保存本地文件。

Alternatives Considered

No response

Additional Context

No response

Implementing this feature in desktop or mobile apps is a piece of cake. However, the web version presents a challenge, especially when it comes to storing files in local storage.

from chatgpt-next-web.

Implementing this feature in desktop or mobile apps is a piece of cake. However, the web version presents a challenge, especially when it comes to storing files in local storage.

Does it mean, in order to ensure consistency across multiple platforms, this project will not consider features that are difficult to implement on some platforms, no matter how useful the feature may be?
I prefer to first integrate these key features on platforms where they are easy to implement.

from chatgpt-next-web.

Problem Description

when writing some personal document I need my own doc, if there are RAG & Emmbedding function to let me upload my personal file it will be very helpful. 当写自己私有文档时需要自己的其他文档作为参考，如果可以提供挂载外部文档的功能将非常有用。

Solution Description

add vector db like chromadb to save local files. 增加向量数据库比如chromadb用来保存本地文件。

Alternatives Considered

No response

Additional Context

No response

Implementing this feature in desktop or mobile apps is a piece of cake. However, the web version presents a challenge, especially when it comes to storing files in local storage.

It is feasible to store only the reference via localstorage. However, when the file changes, the historical references will become invalid.

The most reasonable way is to introduce back-end support and implement relevant object storage capabilities on the back-end

from chatgpt-next-web.

The workload to support embedding seems to be fine, but it needs to be tied to the user's openai account

from chatgpt-next-web.