🇨🇭Switzerland @dan2k3k4

Thanks, fixed!

Released version 1.2.4 of the provider module to map default models - https://www.drupal.org/project/ai_provider_amazeeio/releases/1.2.4 →

MR looks good to me

Tested and confirmed to work in recipes with Marcus

Want to refactor a bit and add tests, so would do this in an MR

Committed hook_uninstall code to 1.2.x and 2.0.x dev branchs

dan2k3k4 → created an issue.

Hey,

Closing this a duplicate of this issue:
https://www.drupal.org/project/ai_provider_amazeeio/issues/3566091 🐛 Key entities created by this module should depend on the module Active

We tried to find a way for the created keys (via Key module) to automatically be dependant on the provider module but it didn't work, so we'll need to implement a hook_uninstall to remove those Key entities.

For a quick fix, you can manually delete those keys:

• config/sync/key.key.amazeeio_ai.yml
• config/sync/key.key.amazeeio_ai_database.yml

dan2k3k4 → created an issue.

Merged and updated project page

Released 1.2.3
https://www.drupal.org/project/ai_provider_amazeeio/releases/1.2.3 →

Thanks, merged MR

dan2k3k4 → created an issue. See original summary → .

Cool, thanks, merged MR

Yeah, I think using hook_uninstall is the way to go

Cool, many thanks!

Merged the MR

Thanks, merged MR, and marking as fixed

Thanks, merging MR and marking as resolved

Thanks, merged MR and resolving this issue

Thanks! Updated the branch default, marking as fixed

Merged to `2.0.x` and also `1.2.x`

We will address the recipes issue in the follow-up
https://www.drupal.org/project/ai_provider_amazeeio/issues/3566345 ✨ Create a config action for provisioning trial access Active

And/or likely try the non-config approach mentioned:
https://www.drupal.org/project/ai_provider_amazeeio/issues/3566117#comme... 📌 Anonymous free trial on module install Active

Marking this issue as fixed

Thanks!

Merged the MR as the PHPStan issues were fixed and I was able to index content into the VectorDB using AI Search.

We can go through adapting / adjusting any other code relating to the VectorDB provider part in follow up issues.

Marking this as fixed.

dan2k3k4 → made their first commit to this issue’s fork.

dan2k3k4 → created an issue.

dan2k3k4 → created an issue.

Pushed a commit that should fix it:

1.2.x - https://git.drupalcode.org/project/ai_provider_amazeeio/-/commit/173de57...

2.0.x - https://git.drupalcode.org/project/ai_provider_amazeeio/-/commit/8665935...

In short, added this to the keys: $key and also $database_key

          ->set('dependencies', [
            'module' => [
              'ai_provider_amazeeio',
            ],
          ])

Would it make sense to link to the usernames instead of just breidert and Wouters_f as text? "non-Drupal" community people might not know what that's for.
Although it seems like it was only done for those two people but no one else, so perhaps remove it ?
To keep all names as just names without d.o (username) in?

Add Speaker: to the 2nd workshop for Vincenzo
Likewise, add Speakers: to the webinar of Dries/James
The AI Agents in Drupal CMS - Create your own agent one should say Speaker and not Speakers as only Vincenzo listed

Oops, this issue slipped away.

I've pushed the following commit to reduce the README.md file

https://git.drupalcode.org/project/ai_provider_amazeeio/-/commit/d3fe891...

Changes to the backend for the amazee.ai provider are already in production.

Changes for the AI provider module are in an MR and going through some testing with other developers:

https://git.drupalcode.org/project/ai_provider_amazeeio/-/merge_requests...

The way it works is that it runs a POST request on module install to generate a new free anonymous trial account. This is then stored in the module config settings just like it does when logging in via email.

A small message appears on the settings page mentioning that you're on an anonymous account with limited budget. If the budget is reached, an updated AI Quota Exception is thrown mentioning to create an account with a link to the provider module settings page.

dan2k3k4 → created an issue.

dan2k3k4 → created an issue.

Hey Mike,

I assume you still haven't been able to reproduce yet?

Perhaps we can close this issue?

Hi

The use of the recipe(s) is not required, the module works out of the box (once you've registered / authenticated your email) without needing to use a recipe to configure it.

Is there anything else needed for this issue, or should we close it?

Adding some notes

Made a quick video edit

https://youtu.be/1kzlxrM17d0?si=PCs_gapumHJ1eNPx

Looks good to me

Thanks, looks good to me.

We might want to refactor a bit to separate out the calls to the API Client, as it is being used for both API and LLM calling.

Regarding your other questions:

• Billing - this is set to a fixed amount per month, depending on the level you select (you'll see this in the Drupal provider page in ~30 days)
• Usage - we currently don't expose this to the provider settings page

We plan to improve the experience and UI regarding this on the provider settings page, ideally to make it as "painless" as possible such that you don't need to worry about logging into another place to manage things.

Feel free to get in touch with our product team, we'd love to discuss what you're up to and how we can help with your needs:
https://amazee.ai/contact-drupal-ai

With the latest version of the provider module, we're no longer reading the amazee_host value from config but directly from the module code itself:
src/AmazeeIoApi/AmazeeClient.php?ref_type=heads#L23

Which version of the provider module are you using?
Did you use a recipe to install it / set it up? I might need to bump the versions of the recipes.

Hmm, which region and model are you using?

And what's the prompt or any other settings?

You may have a malformed prompt, which might likely be an issue with the AI core and not the provider module.

If you could retry and note the rough timestamp, then I can let the team know to check the server logs around that time for the specific region.

Hmm, will need to check backend API logs with the team.

It might be that it's expecting a system prompt to be set and/or a user prompt to be set. You can test a bit with the API Chat Explorer module, which also gives example code.

This would likely need some backend (api.amazee.ai) changes too

Hmm, you may need to run drush updb or potentially drush deploy (assuming you have config-export before running deploy, as that runs a config-import).

Or if you run drush config-export

You can check the config settings of the module at:
config/sync/ai_provider_amazeeio.settings.yml

The key amazee_host should be set to the URL: https://api.amazee.ai

Then if you make a change to the config file, make sure to config-import again either by drush config-import or by drush deploy

Thanks, looks good to me!

Looks good to me, thanks!

Merged

Merged

Merged

Fixed

Yeah, I haven't figured out how to do that, if it's only the "owner" who can do it?

Merged

Merged

Marking as resolved, the issue was about having an old key in the Key module but using a different host name.

Adapted code to read the value of `AMAZEE_API_HOST`

`const AMAZEE_API_HOST = 'https://api.amazee.ai';`

Changed the way the login state is shown by checking for existing LLM key and existing VectorDB key. It does not check if they are valid though.

ah yes, since the module was built without a README.md, I took the project page and used it as a basis, I'll clean that up too once I've finished another bigger task.