Feature request - Add custom headers to the AI assistant

[pythoninja]

My feature request is about headers:

1. Add a User-Agent header to the AI assistant and/or
2. Add custom headers like Open WebUI (for example) did (check the screenshot).

I am using a LiteLLM instance to route the request to the right model, and currently I want to mark a Poznote's request with the custom header x-litellm-customer-id.

LiteLLM logs:

Open WebUI headers configuration:

[timothepoznanski]

Thanks for the suggestion, both parts are implemented and will ship in the next release.

Every request Poznote sends to the AI server now carries a User-Agent of the form "Poznote/<version>", so your LiteLLM logs will show where the call comes from without any configuration.

There is also a new "Custom headers" field in the AI Assistant settings, both in the instance configuration and in the personal one. You write one header per line as "Name: value", for example "x-litellm-customer-id: poznote", and they are sent with every request to the server, the connection check included. The field is available for Ollama, LM Studio and custom URLs.

Since such headers often hold credentials for a gateway, their values are stored encrypted and shown masked once saved. You can leave a line untouched to keep its value, retype it to change it, or delete the line to remove the header.

A few headers are refused: the ones that frame the request (Host, Content-Type, Accept and the like) and Authorization, which stays with the API key field. You can set your own User-Agent line if you prefer it to the default one.

Let me know if this covers your LiteLLM setup once you have tried it.

[pythoninja]

Let me know if this covers your LiteLLM setup once you have tried it.

Thank you, I will report back when I test it.