this is the second error I've found. Maybe I've got a formatting error but I tried to cover all my bases before posting. This happens on macOS as well as on iOS.

Best Raphael

Hi, Since the last update I keep getting an error. It is going through to OpenAI as I am being charged but the text isn't displaying on the obsidian page. It was working okay before I updated the plugin.

Along the same lines of #23 Text Generator: Template Packages Manager uses alt + 3 by default. This is how you type a # on a mac obviously this is used a lot when writing markdown.

Everytime I try to use the plugin lately I get error 429 (too many requests) but I'm not sending many requests at all

getLine was called on line 0
plugin:obsidian-textgenerator-plugin:200 {prompt: 'Explain the similarities and differences between quantitative and qualitative variables', max_tokens: 160, temperature: 0.7, frequency_penalty: 0.5}
plugin:obsidian-textgenerator-plugin:93 Error: Request failed, status 429
    at new t (app.js:1:699529)
    at nw (app.js:1:699721)
    at app.js:1:700398
    at app.js:1:235201
    at Object.next (app.js:1:235306)
    at a (app.js:1:234045)

Obsidian 1.03 on MacOS

Steps:

• Open Command Palette with Command+P
• Type "Text" to highlight the Command "Text Generator: Set max_tokens"
• Press "Enter" Button

Expected:

• Popup Dialog shows to edit the max_tokens

Observed:

• Popup Dialog flashes for some milliseconds ;)
• (probably) saves the existing setting
• shows a toast in the upper right corner

Remark: Option/Alt+1 opens the Popup Dialog without Problems

Thanks for the great plugin & BR Günther

Hello, in v1.1 there was an error message when pressing Ctrl+J or executing the command from the sidebar. Shortly after you released v1.2 to fix it but the error is still there :(

I installed the latest version of the plugin, I've updated Obsidian to the latest version. I tried one of your examples with the frontmatter. When I want to generate text, the statusbar hangs on Text Generator(160):processing... I restarted the app a few times but no luck. I also made a brand new vault without other plugins but still the same behaviour. Is there a way to debug this? I am a bit familiair with the developers tools and network window, but I'm not quite sure where to pay attention to.

Thanks

Hi,

If I try to use any frontmatter input, I always get the error message "Please provide a valid frontmatter!"

I tried to open a fresh vault (obsidian version 0.14.15) with only the textgenerator plugin installed (version 0.0.8), but the error persisted. The used input used was just:

---
title: "How to write a youtube video"
---

Brain storming ideas 

Generating text with just selecting it, works fine.

Hey there! Thank you for putting this plugin together.

Since I mostly use my own trained models, could you possibly add an option to select (or simply provide a textfield for the name) for using a custom model?

That would be amazing! Thank you.

It appears that there's a bug where the downloaded Prompt templates are not installed correctly if a different prompt template path is set in Options.

Steps to reproduce:

1. Install and enable plugin
2. Go to Options and change "Prompts Templates Path" to a different path (e.g., "openai/prompts")
3. Use Ctrl+P and open the "Template Package Manager"
4. Install any of the Prompt packages
5. Try to insert a Template using the "Insert Template" command

Actual Result:

• The "Insert Template" pop-up shows an empty list.

The templates are installed to the default path: "textgenerator/prompts, instead of the path set in the Options.

Please let me know if you need more details.

Thank you for this wonderful plugin! William

With the new version, cmd+Q seems to be used as default hotkey. This should be avoided, since cmd+Q is the default macOS keybinding for quitting an app. (equivalent to alt+f4 on windows)

Might be a bit premature since it looks like templates are not fully out yet, apologies if some of these are already planned. Some of these are probably bugs.

Escapes and text replacements:

1. custom escape characters, through helper function
2. substring helper function, and corresponding documentation :P
3. text search and replace
4. text search and index

Children: 5. avoiding overlap for children; currently the same child can be included multiple times in the children array 6. *option to avoid self-child; currently the parent note can be included in the children array 7. option to generate headings for children 8. including children of children; currently child notes of children are inaccessible

*has workaround using #if, but imo having this option is better

Links and Embeds: 9. option to exclude headers when including header contents 10. including referenced blocks; the contents of blocks of the form [[note#^blockref]] seem to not be included anywhere 11. option to include or exclude referenced blocks from children; 12. option to replace embedded internal links with their contents; of the form ![[note]] or ![[note#^blockref]] 13. *option to remove or replace embedded internal links to non-notes; for instance ![[image.png]]

*has workaround if text search and replace or text search and index + substring are implemented, but cumbersome

Quality of Life & misc: 14. option to exclusively use {{headings.heading}} instead of {{heading}}, to ensure consistency (eg for a heading called "context") 15. *option to show multiple responses to pick and choose between them; ie the n parameter from OpenAI's Completion API 16. *option to retrieve and view logprobs from OpenAI, as well as the tokens 17. insert template to clipboard

*probably a lot of work

The Part where I Write a Lot (More):

Hi! I'm loving this plugin so far, and really appreciate what you're doing! I'm loving the concept of templates so far, but a lot of these features will make it more practical for my use case. I hope you'll consider implementing some of these. Here I'll try to explain motivations and/or vague potential implementation features:

1: default is HTML escaping, but to my knowledge OpenAI works with JSON(L) format, and I'm not sure if escaping for JSON currently exists? can replace with a simple helper for JSON escaping, but I suggested a search-and-replace since it could work for all potential future AI-as-a-service APIs 8: potentially a field in the plugin options, but also consider writing it as a helper function, although I don't know if helpers are allowed to have side effects. This helper function takes an integer max_depth 9: again, potentially a toggle in the plugin options, but would be nice to be a helper function instead? or can be a {{headings.heading.title}} and {{headings.heading.content}}, but that seems unnecessarily verbose 10: consider a references that is separate from children. Also consider generating references for children (11). 12, 13: could group .txt under 12 instead of 13, although that's probably unimportant for most people 14: please 15, 16: this would imply the creation of a temporary file at best, or at worst a custom popup/window; additionally for the vast majority of people this is probably completely irrelevant. Personally, seeing logprobs for longform text allows me to read into the intent of GPT, and sometimes good information can be found by looking at what GPT would have said. Still, not exactly the most commonly used feature, and (17) is a good enough replacement where I can just paste into the OpenAI playground.

Given the Obsidian context, I think it would be awesome to be able to select the editing endpoints/models (potentially with a different command?)

The edits endpoint can be used to edit text, rather than just completing it. You provide some text and an instruction for how to modify it, and the text-davinci-edit-001 model will attempt to edit it accordingly. This is a natural interface for translating, editing, and tweaking text. This is also useful for refactoring and working with code. Visit our code guide to learn more. During this initial beta period, usage of the edits endpoint is free.

Exploring the plugin, I found that I frequently switch between two token amounts: 100 tokens for finishing a sentence, and 500 tokens for writing a short paragraph.

Right now, that means frequently using the "change max tokens" command. I think this could be streamlined by being able have commands/hotkeys for each separate token amounts. For example:

• Command 1: Generate Text (with 100 tokens)
• Command 2: generate text (with 500 tokens)

and then I simply have two hotkeys, instead of constantly changing the token amount between the two values.

right now, when the cursor is in the middle of the line and I trigger "generate text", it enters the text right at the cursor position, although it should be appended to the line instead.

(also, and this is mostly minor, when appending, and the last character is not a space, it should add a space, so that you don't always to do that manually.)

First of all, thank you so much for the plugin!

Consider the following three scenarios:

1. I have some text, and want the AI to complete the text, for example finishing a sentence I started.
2. I have an instruction, and want the AI to write text based on that prompt, for example "outline a blogpost about x"
3. I have text and an I have an instruction, for example "rewrite that sentence, but in simpler language"

How exactly do I use the plugin for these different scenarios? Experimenting with the plugin a bit, it seems that even the basic "generate text" command covers all three cases, depending on how I phrase the input text. But to me, it feels like having three commands for the three different scenarios would make this far more intuitive and also more clear in case of what happens in scenario 3.