The @ lets you selected un supported assistants and throws an error

[AFaulkner1990]

What happened?

We have two versions of LibreChat running, one as a general user site that uses ModelSpecs and another power user "admin" site, that lets users create assistants/agents etc.

My assitants config looks like this:

  azureAssistants:
    capabilities: ["code_interpreter", "actions", "tools", "image_vision", "retrieval"]
    disableBuilder: true # Disable Assistants Builder Interface by setting to true
    privateAssistants: false # Enable private assistants by setting to true
    supportedIds: ["asst_xxxxxxxxxxxxx"]
    pollIntervalMs: 3000 # Polling interval for checking assistant updates
    timeoutMs: 180000 # Timeout for assistant operations

"Aila code" is the model spec using azures assistants asst_xxxxxxxxxxxxx

However when I type @ as suggested in the chat input box, it shows all the other assistants, which it then obviously fails if you try and use.

Looking at the code this only seems to be for Admin accounts, I am unsure what you want the expected behaviour to be so have not done a fix

Steps to Reproduce

• Login as an Admin
• Create a couple assistants
• Add only one of them to your supportedIds option
• Create a ModelSpec using this Id and set disableBuilder to false
• Type @ into chat input box and see if other assistants show

What browsers are you seeing the problem on?

Chrome

Relevant log output

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

After further review, realized this is not a bug and rather expected behavior. The admin should indeed see all assistants.

[rba100]

@danny-avila this also affects USER role with respect to the endpoints that the modelSpec's refer to. E.g. in the above scenario whilst a USER could not see the assistants, they would see the gpt-4o endpoint that powers the modelSpec. The expected behaviour is that USERs should only see 'addressable' models in @ mentions, in line with what they could select from the main model drop down.

I think this should go back to being a bug. Should it be re-raised with the specific scenario or is this good enough?

Version 0.76 (librechat-dev:6c33dc2eb37a152334ae5fc2bb2e84c13c7cb311)
endpoint: Azure OpenAI chat completions endpoint
modelSpec: non–assistant-based, points to basic azure chat completions endpoint.

[danny-avila]

@rba100 you can open a new issue, describing this specific scenario.