Search

[moloch--]

Search functionality in the web UI and API

[areknawo]

There are 2 types of "search" I have planned:

• find & replace right in the editor (inspired by similar features in code editors like VS Code, with support for regex, operating only on content available in the editor;
• general search (2 stages):
  1. initially querying only data like description, title, and other metadata, together with a command palette for faster navigation in the entire app;
  2. full-text/vector search for querying through the entire available dataset (metadata & content); it's the end goal for a great search, but it will take the most time to implement;

Let me know which type of search you were referring to, so I can prioritize its development.

[moloch--]

General Search / (ii) full text is the biggest thing missing for what we'd like to use vrite for, with (i) being a close second.

[areknawo]

Makes sense. (i) is coming in the next week or two, as it's simpler to implement. (ii) will come later on (likely late July/early August).

[areknawo]

@moloch-- Took a little longer than planned, but a search is now available in v2023.8.18-beta with the addition of a command palette. You can both search through your content and ask questions (powered by hybrid vector search and GPT-3.5). search() method is already available in the SDK with the same capabilities as the built-in search, while additional /search/ask endpoint is also available with SSE streaming. Docs for both and SDK support for the /search/ask endpoint will be added soon.

[saiqulhaq]

I am using https://app.vrite.io/
I am having an issue to add my OpenAI access token
Could you share what is the cause?
Should I use self-hosted version to get make OpenAI extension works?

[areknawo]

In the hosted version you don't supply the token.

The AI/semantic search is built-in and enabled by default. You use the search bar in the dashboard or a ⌘K/Ctrl K shortcut to open the palette.

On the other hand, the GPT-3.5 extension integrates the AI into the editor. Once installed, you can use the new "Block Action" button (on the right side of the editor) to use GPT-3.5 to generate content right into Vrite Editor.

If you want to self-host then you can choose to enable the AI/semantic search by providing your OpenAI token as an env variable.

However, as extensions aren't currently supported while self-hosting, the GPT-3.5 editor integration is only available in the "Vrite Cloud".

[saiqulhaq]

Oh, got it
I found two issues:

1. chatgpt "block action" button is missing after I click the "generated button". I think I have to wait a few minutes, then it will appear. Any reasons for this? is there a rate limiter?
2. the extension doesn't read the article I wrote in the editor. so it has no information about what I added to the editor. the chatgpt is it expected to be used for generating articles using their own DB only? can't be used for Question Answer? FYI I have 1 article only

[areknawo]

1. This might be a bug. I will open an issue for that and work on fixing it.
2. Right now, the "block action only uses the GPT knowledge and optionally the active paragraph (when Include context is used) to create new content.

If you want to generate an answer from your existing content base, you can use the Ask option in the command palette. Here's both the search and AI "Q&A" in use:

So, in general:

• Currently, the extension is for generating new content (e.g. "Write an introduction to JavaScript") or editing paragraphs (e.g. "Make it shorter")
• The semantic search, AI-based question-answering (for your content base), and a command menu are all in the command palette and accessible via API.

[saiqulhaq]

Thanks for the answer
I can use "Ask" feature without any issues
By the way, what is the rate limit of "Ask" feature?

[areknawo]

Currently, for the API, it's 500 per 1 min.