Hello everybody! Nice to meet you, my name is Marcos and I'm a data engineer in Brazil I work at a large company where we have an internal AI gateway (LiteLLM-based) with 130+ models available. I'm deploying the OpenHands GUI internally so our analysts can build apps and tools using our internal data, behind VPN/VPC I got it running successfully with our gateway using Docker (

docker run

with

--network host

+

SANDBOX_USE_HOST_NETWORK=true

to solve VPN routing issues) My question: is there any plan to support multi-model configuration via the GUI/settings.json when running via Docker? I see that

config.toml

already supports assigning different LLMs per agent (e.g.

[llm.code-gen]

+

[agent.CodeWriterAgent].llm_config

), but this only works in development mode. It would be very useful to set, for example, o3 for planning, Claude Sonnet for coding, and a cheaper model for repo exploration — all from the GUI Thanks!

Yes, @modern-fireman-99755 and I are working on it, hoping to get a PR in very soon!

Amazing! Thank you so much. Looking forward to testing it out. If you need any feedback from the enterprise/VPN deployment perspective, feel free to reach out!

Yeah, maybe @hundreds-twilight-32833 would be interested in collecting feedback?

Anytime!

Happy to chat @ambitious-microphone-55463! Feel free to schedule some time https://calendar.app.google/YJ2ZeEWd9fexRUhQ7