Hi

First, thanks for your great work !
I met the following issue which is new after recent update (just updated with BRAT).

When tab is on main frame, all is fine, but when I the tab is at top right corner I have no access to the text input form.

Hope it's sufficient to investigate

Thank you for the plug-in which is nicely integrated into Obsidian.

At the end of most chats is a recommendation "As for tags related to this topic..."

The tags do not seem to be from my vault. What do you recommend we do with them? Is it suggested that we create them for our vault?

It's great to have a plug that is integrated well into Obsidian

When I try to start a chat, I get this response:

I'm sorry, but I am unable to generate a response at this time. This may be because your request was cancelled, GPT4 is currently in use, or an error has occurred. Please check your settings and try again later.

I disabled Text Generator and re-started, in case this was a problem. I'm using the same API. It works with Text Generator.

Here's what the console says:

The engine is GPT3.5 Turbo.

Let me know if there's anything you'd like me to explore further.

It would be great if obsidian displays the text being generated in real time while GPT is generating the text. This allows me to read the message as it is being generated and helps me decide whether I should stop GPT from generating more text.

Currently, the respond box is only loaded after the entire response message is completed. As a user, I feel a little uncomfortable waiting for the message to be completed. I would like to read it while it is being generated.

I like to ask ChatGPT to describe something, then copy its response into a note. Right now, I have to manually highlight the text I want to copy, paste it, and manually clean up the formatting.

It would be great if there is a button that copies everything in a response box into clipboard. I can then paste it anywhere I want and modify it.

Hello!
I would like to try the release, using BRAT, but no release are found.
I don't know which branch or tag to use to build the main.js, so i'm little lost.

Since I saw you put latex in the system role prompt, my first thought was to test my latest chatgpt hobby: text adventure gaming with dynamic diagrams!

It did not quite work and I'm not sure why. Here's the prompt:

what's the relationship between the sun and its planets? Draw in a mermaid diagram code block

I'll paste an image of the output so you can see that it is overflowing wayyyyy off the screen

And the text:

The relationship between the sun and its planets can be represented using a mermaid diagram in Obsidian. Here is an example code block that illustrates this relationship:

```mermaid
graph LR
    A[Sun] --> B[Mercury]
    A --> C[Venus]
    A --> D[Earth]
    A --> E[Mars]
    A --> F[Jupiter]
    A --> G[Saturn]
    A --> H[Uranus]
    A --> I[Neptune]

This code block uses Mermaid syntax to create a graph with the Sun at the center and each planet connected to it by an arrow. The arrows represent the gravitational pull of the Sun on each planet, which keeps them in orbit around it.

Note that you will need to have the Mermaid plugin installed in Obsidian to render this code block properly. You can install the plugin by going to the Community Plugins tab in the Obsidian settings menu and searching for "Mermaid".

Any clue what to change here to make it work?

update: looks like it just didn't close the code block. Let me see if I can get the bot to do it w/o code changes.

As of Obsidian 1.0 interactive elements should use the var(--cursor) variable, which renders as the default cursor. The pointer cursor should only be used for links.

See
https://forum.obsidian.md/t/1-0-theme-migration-guide/42537#cursor-12

Congrats on launching this plugin, it's fantastic!

It can be useful to configure how many previous messages in a conversation are sent to the API. In some scenarios, like simple language translation, the history of the conversation does not add any value, but does increase the token count.

In some other clients I see you can set the number of previous messages (like 0 to 100) to control how many are sent.

Related to this is the ability to delete various messages from a conversation so as not to resend something that in the end wasn't valuable in the conversation.

Hi, this is definitely an awesome plug-in to see! Thank you for making it.

I'm wondering if we can improve the markdown rendering of the responses in the following ways:

1. Support math rendering. I see you using remark/react-markdown to render the dialogue bubbles, I'm sure it supports a go-to math rendering plug-in (check this). It would be very helpful to people who work in academia to have latex rendered in response!

2. Rely more on obsidian's rendering tools

   1. I think for code blocks rendering, obsidian exposes its global Prism object through loadPrism. I'm not an expert on this and not sure if it can be helpful or not, so this is just FYI.
   2. Similarly, I'm wondering if there is an API to utilize obsidian's Markdown engine so we can throw remark/react-markdown away, with way better consistency with the rest of the app. However, after a quick view of the relevant APIs, I'm not able to come up with a quick way to do this in our React components. It will be great if anyone else with more knowledge can help :)

Hey! I'm getting this error message. I provided my api key from openAI. and the model i have selected in the plugin settings menu is 3.5, not turbo, or 4. Do you have an idea of what could be causing this issue?

Hi, @vasilecampeanu !

While GPT-* models are awesome, there's an issue of privacy arising from using an external third party API, especially when the information may be sensitive. More than that, they don't allow as much customization for one's needs as the local models. In my mind, adding support of local models is one of the top priorities of such a plugin.

Fortunately, thanks for people like oobabooga making a wrapper for many possible local models like LLaMa, Vicuna, MPT, Galactica, etc it's pretty straightforward to implement. Just replace the call for OpenAI's API to oobabooga API's endpoint. For example, how I did this rizerphe/obsidian-companion#13 (may require some tweaks)

supports streaming mode as well

I configed openai key.but, aways get this page.

Currently we can globally configure the

• gpt API (ex: 3.5, 4.)
• Prompt
• max tokens
• and so on

It would be super helpful to be able to set the parameters at an individual conversation level. Therefore, if they are not set at the conversation level, the conversation uses the defaults defined in the settings.

If they are defined in the conversation, then they use those specific settings. This can help tweak certain conversations for very specific purposes. For example, I might have one conversation where I want to use the 4.0 API, along with a very specific system role prompt, while another conversation would use the 3.5 API with a very different system role prompt.

4.0 API works better for certain things, especially since it is more sensitive to the system role, whereas 3.5 is good for many things and cheaper.

I know this is a big ask, since each conversation would need to maintain its own configuration and UI for it.

Using a key links all your searches and responses to an identity. Requests will always be linked with an IP address, but that address is shared in most cases. I'd love to use the plugin but don't want to create a key and add my identity to big metadata bucket in the sky. Any chance of modifying the plugin to not use an API key, even at the loss of some functionality?

Is there a way to adjust the font to a larger size?
I'm using Minimal theme together with Minimal Theme Settings and Style settings plugins
It's incredibly tiny right now, and any font size settings in obsidian and these plugins doesn't affect it.

Since installation and with different models, the plugin keeps returning an error. Here is the return on the chat view and the console. I'm on Obsidian 1.2.2 installer 1.1.9, MacOs Ventura

I am getting this error after a fresh install of this plugin.

I have all the settings to the default except my api key. Any help would be very much appreciated.

After 15 minutes of using ChatGPT it responds with "Error: HTTP error! status: 400"

There is a workaround where you disable the context feature (the green button in the upper right). Still, it would be nice if I could keep this feature.

I tried to install the plugin on Obsidian mobile through BRAT, but after BRAT installs it, I cannot enable it in Obsidian community plugin menu.

I pressed the enable toggle button, but it refused to toggle on.

I don't see any error notification, and I don't know how to get the logs to provide you more information.

The only community plugins I have in my vault is BRAT and Weaver (this).

I can install the plugin manually or via BRAT, but cannot enable it despite restarting Obsidian several times. All other community plugins can be enabled/disabled without problems.

Displaying the content generated in real-time rather than waiting for the whole content can help user to interact with the plug-in more efficiently and save much time in some cases, which feature is provided in the similar plug-in https://github.com/logancyang/obsidian-copilot.
Leaving the uncomplete content unformated during the generation is just okay I personally think.

First, Kudo's. Loving the styling of this plugin.

When clicking on a conversation in the main view, it should open the conversation thread. Currently we have to click on the arrow to open the conversation.

As you can see in the screen grab where I am clicking it seems natural it would open the conversation.