Following is the detail. Unable to generate images
I have a valid API key and able to test @ https://beta.dreamstudio.ai/

Prompt Selection used

A man is playing chess against a robot, the robot is winning and the man is getting frustrated, the robot is taunting the man and the man is getting angry, the man is about to lose the game, 8k resolution, post processing, detailed, robot, chess, man, game, losing, frustrated, angry, taunting, red eyes, grinning face, plans, announcement, dismal, 8k

From command Palette, Select

Obsidian ava: Generate an image based on selected text
Using Hotkeys: Ctrl + Alt + S

Error

Error while generating image
TypeError: Cannot read properties of undefined (reading filePath)

Status bar warning

On hover message

Versions

OS: Windows 10
Obsidian ava Version: 1.0.8
Obsidian version: v1.O.3

Has anyone successfully run the ava api server under windows? I was getting various errors under python 3.9 when trying to start the server. After installing dependencies, trying to start the server yields

File "[PATH]\my-vault\.obsidian\plugins\ava\semantic\env\lib\site-packages\obsidiantools\md_utils.py", line 273, in _get_md_front_matter_and_content
    return {}, file_string
UnboundLocalError: local variable 'file_string' referenced before assignment

Anyone managed to get it to start up?

Feel free to share your Templater templates ❤️.

To be clear Templater is used to quickly explore new features. If these features survive, they can be integrated into obsidian-ava with a more efficient UX.

I would really like to be able to see notes related to the current note in a side panel! Then decide if I want to drag them into the note as links or not. Thanks!

We are currently facing difficulty installing our plugin due to the need of running a Python API on install. This is done through a Node.js file that initiates 'pip install' commands, but we require Virtualenv to be installed in order for it to work. We desire a better solution as there are several issues with relying solely on Virtualenv (since it is not always installed on the host computer, which leads to installation failures.)

We would highly appreciate any suggestions on how to improve the current installation process.

On a fresh install, the Ava program has failed to load properly and all features are broken. I'm running the latest plugin version and I have disabled all other community plugins, my custom css, and created an entirely new vault

Console Log:

app.js:1 Total plugin setup: 753ms
ava: 753ms
plugin:ava:346 Object
plugin:ava:346 e
plugin:ava:346 Ava - Indexing vault with Array(1)
obsidian-ai-c6txy76x2q-uc.a.run.app/v1/search/refresh:1          Failed to load resource: the server responded with a status of 401 ()
plugin:ava:346 Error: Error refreshing semantic search: 
    at Sq (plugin:ava:316:662)
    at async DUt.indexWholeVault (plugin:ava:346:6634)
J3 @ plugin:ava:346
indexWholeVault @ plugin:ava:346
plugin:ava:346 Ava - Unlistening to note events

• Absolute path
• Relative
• Aliases
• ...

https://app.excalidraw.com/l/5MKXLddifTr/9AqOqMCtubu

https://handlebarsjs.com/

1. Completions based on templates
2. You can chain completions (e.g. "this is a nice piece of text [MYPROMPT]" -> "this is bad [MYPROMPT]")

I wanted to express my appreciation for the Obsidian Plugin you created.
The Generate Link function has been extremely useful for me, and I've found myself using it extensively. However, I couldn't help but think how even more productive and efficient my workflow could be if there was a similar function for generating tags as well. I believe it would be a valuable addition and I would greatly appreciate it if you could consider adding it.
Thank you again for your hard work and dedication in creating such a useful tool.

The plugin is affecting global CSS on Obsidian. See the video:

https://www.loom.com/share/366f677059004eccaa418db48afcb006

Automatically link notes after some delay (languagetool/grammarly plugin UX like).

Example:

I have a note "Dog"

In a new note called "Animals" I start writing "Dogs are nice animals", when I start typing "are" it will automatically put [[]] around Dogs (i.e. "[[Dogs]] are nice animals" at the end)

We can use human feedback on AI outputs to improve the models.

Interesting resources

• active learning https://www.argilla.io/
• RL + NLP https://github.com/CarperAI/trlx

The plugin should respect the attachment folder path setting for images generated with stable diffusion.

When I select text and run rewrite, nothing happens. Ditto with write paragraph.

I got the search API and link functionality working, so this does not seem to be an issue with my OpenAI API key or anything like that.

When I look in the console when hitting rewrite, this is what I get:

POST https://obsidian-ai-c6txy76x2q-uc.a.run.app/v1/text/create net::ERR_CONNECTION_REFUSED

stream | @ | plugin:ava:13
-- | -- | --
  | s | @ | plugin:ava:556
  | await in s (async) |   |  
  | search | @ | plugin:ava:552
  | (anonymous) | @ | app.js:1

When I look in the network tab, the response section says: "Failed to load response data: No data found for resource with given identifier."

Basically the same thing happens with Write Paragraph.

I'd be happy to provide you with any additional information that you need.

Currently, the env would be installed within the plugin folder and would be synced by Obsidian Sync. It would be great if the env can be installed outside the plugin directory, and one can specify the location of the env in the plugin settings.

Hello, I am a user from China, when I register with open ai, my cell phone number cannot be verified.

Similar to rewind.ai, extract text from images and index it.

https://github.com/louis030195/obsidian-ava/blob/5386875803221e2655bb122b3c6387c517e4fb5a/semantic/.gitignore#L1

Otherwise, for git obsidian user it tries to commit this env with huge files, maybe ml models etc.

https://github.com/MubertAI/Mubert-Text-to-Music

https://docs.langa.me/

i.e. get intellectual feedback and reverse engineer your brain

• mono modality (image)
• multimodal (text - image) (image - others)

There is only a Sign in with Google option. Can we add email or something that doesn't link to a third party service?

Hi @louis030195 and @hotkartoffel. We (the Obsidian team) noticed that AVA has client-side telemetry code posting user data to PostHog from directly within Obsidian. We are still drafting a privacy and data collection policy for plugins, but there are a few guidelines we have already agreed on. Mainly, plugins should not submit user data to third-party telemetry services from directly within Obsidian, even if the data collection is opt-in.

Client-side libraries like PostHog's make it too easy for user data unrelated to your plugin to get collected. Instead, we ask that you move that data collection to your backend service; you should still be able to collect the same relevant usage information without risking user privacy.

A sidebar the shows the most similar notes would be awesome!

Even though I copied it from openai.com, it does not recognize my key:
Not sure, do I need to regard something else?

Hi,

I added these values to Settings:

When I press "Save", I noticed this error in the console:

When I select a text and use "Generate an image based on selected text" nothing happens.

Any ideas?

The state of the vault's embeddings is fixed in time, requiring a reboot of the API to have the "search" features updated.

mfarragher/obsidiantools#23

Would be good to have prefix settings for suggested tags

For example:

aka my prefix setting is ss\
all suggested tags would be ss\blah1 ss\blah2

When trying to start the search API, it would silently fail. When I looked at the console, it was giving this error:

node:fs:1536 Uncaught (in promise) Error: ENOENT: no such file or directory, stat '/Users/danthompson/Obsidian/Synced Vault/.obsidian/plugins/ava/semantic'

I figured that this error was due to it not automatically creating the right directory. I double checked that the issue wasn't due to an old install (I am always searching for obsidian AI plugins so I installed Ava back when it first hit the store). I uninstalled and reinstalled, but it did not help.

Then I manually created the directory semantic at ~/Obsidian/Synced Vault/.obsidian/plugins/ava and running start search API worked.

So I think you need to add a mkdir step, at least when the plugin is installed through the Community Plugin interface.

If you want mobile compatibility, comment how would you use it, please?

Some users requested the ability to customize the behavior on inserting links. For example with a template.

Example: comma separated list instead of bullet list.

The new canvas feature is pretty amazing. It would be cool if you could right-click on any card and click convert-to-image and it would convert the text description to a stable diffusion, dall-e or midjourney image.

If I'm not actively using Ava, it would be nice to be able to stop the search API so that it is not running all the time in the background. Especially since it generates a ton of notifications. But also because I assume it is updating the embeddings constantly, which has a cost.

Of course, it's possible to simply disable Ava, but that is more time consuming than a simple command to disable the search API.

Waiting obsidianmd/obsidian-releases#1211

Workaround in the meantime: git clone the repository in vault/.obsidian/plugins

Re: Excalibrain OpenAI Script: zsviczian/obsidian-excalidraw-plugin#833

Re: Galaxy Browser: https://www.youtube.com/watch?v=8Y7_gvmMyKY

Hi,
the "Get Wikipedia Suggestions" is an interesting feature.
But:

• a new Side Panel is created for the links. Wouldnt it be better to add the links at the position of the cursor in the current document?
• Also I can't mark the links to copy them for later use.
• You mention "Press F1", It's not working on macos (its Command + P there)

Maybe I miss a point here. Please advice or fix.

Nice work so far & BR Günther

Maybe another repo with community's templates that you can clone in your vault & sync regularly

here oneliner to convert
https://gist.github.com/SheldonWangRJT/8d3f44a35c8d1386a396b9b49b43c385

It seems the documentation is not good enough because I see people struggling to install the plugin, suggesting a Loom end-to-end, especially Templater configuration is difficult to understand at first.

https://platform.stability.ai/docs/features

https://www.midjourney.com/home/#about
https://beta.dreamstudio.ai/dream

Hi, I have tried the plugin and I like it, but I have some concerns about privacy.
When you rewrite the text or activate the indexing of the notes to link them.

What happens to the data?

Thanks

The settings interface of the plugin will not accept OpenAI API keys:

Is there some kind of workaround? Or a way of understanding what the plugin is expecting of the API key?

I always try to connect my notes together using links or tags.

We can automate it using semantic search https://github.com/louis030195/obsidian-search

TODOs:

• metrics / evaluation to know if fine-tuning is worthwhile
• check if there is an easy way to fine-tune with supervised data collected through user interaction

This is important information that would influence whether we try to implement inference in JS directly (ONNX, TFJS, etc.) which makes it more difficult to fine-tune.

Should pick last model by default for both Stable Diffusion and GPT3 (use engine API)

https://www.aleph-alpha.com/