jcollingj / caret

Caret, an Obsidian Plugin
https://caretplugin.ai/
MIT License
101 stars 9 forks source link

The part about adding a custom model is unclear. #18

Open jkjoker opened 3 hours ago

jkjoker commented 3 hours ago

I really like this plugin, especially the concept of having conversations within a canvas, which is full of imagination. However, I encountered a problem when configuring the model.

Here's the process: Usually, to configure a large model provided by ollama, you need to provide the model name and base URL. The model I want to add is qwen2.5:3b.

Here's what I filled in when configuring: image

It didn't work when I used it, and the following prompt appeared: image

I think it's because the parameters are not filled in correctly, but I don't know exactly how to configure it. I would be very grateful if you could give me some suggestions!

(ps: Perhaps the model configuration part of this plugin is too complex. For example, for ollama, the ideal situation would be: as long as ollama is selected, the plugin can automatically detect which models are installed locally and available, and you can see a list of models. Then, when switching models, you can simply check the box.)

Latest update: Hey, I just posted about a bug I found, and guess what? I tried again, and it's working now! Cool! Anyway, I'll keep this issue here and share some other thoughts that might help.

Here's what I've been thinking:

1、I'm still a bit confused about when and how the workflow I set up actually gets used. Can you explain that? 2、The chat history is saved as an XML file, which is a bit hard to read. I think it would be better if it was saved as a Markdown file. All we really need is a timestamp, who said what (user or chatbot), right?

jcollingj commented 2 hours ago

Howdy @jkjoker I am sorry but I am limited to only english. Is there any way you could repost this in English?

jkjoker commented 2 hours ago

Howdy @jkjoker I am sorry but I am limited to only english. Is there any way you could repost this in English?

I apologize for any misunderstandings that may have arisen due to the language barrier. To ensure clarity, I've switched to English.