Open-Webui Migration

[spammenotinoz]

Hi,
I am considering a move from Open-WebUI primarily for the slick UI and Code Interpreter wondering the feedback from others as there is limited feedback on forums;

• Is the Key for code interpreter set per user, as I couldn't see where to define it Globally in the .env?
• How do you deal with the lack of Web Search? Seems limited to Agents, which seems clunky.
• For custom endpoints, can you enable "Anthropic Prompt Caching" and can you set "Icons". I know you can set a base_url, but mine is in OpenAI compatible format, so seems like I need to define as Custom, but then loose Anthropic Prompt Caching (which my provider supports)
• Does this project support the Beta LLMs from Google that let you create\edit images. This has been popular on Open-WebUI?
• Is there an equivalent of "Call Support"?

Not interested in migrating data, but any feedback from a user experience would be appreciated.

[danny-avila]

Thanks for your questions, happy to answer!

Is the Key for code interpreter set per user, as I couldn't see where to define it Globally in the .env?

Absolutely, this is in the documentation (though we need to make it more apparent):

https://www.librechat.ai/docs/features/code_interpreter#in-librechat

a. Per-user setup: input your API key in LibreChat when prompted (using the above methods)
b. Global setup: use LIBRECHAT_CODE_API_KEY environment variable in the .env file of your project (provides access to all users)

How do you deal with the lack of Web Search? Seems limited to Agents, which seems clunky.

Something great here is planned in the short term, stay tuned, not just limited to agents.

For custom endpoints, can you enable "Anthropic Prompt Caching" and can you set "Icons". I know you can set a base_url, but mine is in OpenAI compatible format, so seems like I need to define as Custom, but then loose Anthropic Prompt Caching (which my provider supports)

Unfortunately, no, Prompt Caching is not a toggle as custom endpoints are mainly "OpenAI-compliant" and not "Anthropic", but something here is due implementing for other reasons as well.

Does this project support the Beta LLMs from Google that let you create\edit images. This has been popular on Open-WebUI?

Not yet but is planned.

Is there an equivalent of "Call Support"?

Not yet, mainly because realtime audio is not widely supported in a multi-user context yet, and solutions combining STT/TTS can be quite cumbersome and prone to latency issues.

[spammenotinoz]

Thank-you, appreciate the quick response. Despite skimming and searching the docs, completely missed the LIBRECHAT_CODE_API_KEY, guess I was searching for INTERPRETER or something.
Since your reply, i've done a quick deployment, and impressed how easy to get running \ customisable.
I have a litellm callback that does prompt caching, so I guess it's just the Anthropic Icon I would loose.

From a user perspective, the interface is extremely clean, well-done to the team.