Comparison of Github stats for the four self-hosted LLM UIs I deployed on Kubernetes

#lobe-chat #open-webui #anything-llm #LibreChat

www.githubcompare.com/lobehub/lobe...

Just tested LobeChat as well. Besides the missing Helm Chart, everything can be configured easily via EnvVars - even plugin settings (for example Google Search via CSE). Its also very feature rich. Only downside that it currently does not support Google OIDC auth natively. And >50k ⭐ like Open WebUI

Same goes for AnythingLLM: the integration of all the different vector dbs, embedding / transcription / Voice & TTS providers is nice - but you need to configure everything in the UI. Also no Helm Chart available at all - and the provided K8s manifests use a Deployment & PV instead of a StatefulSet.

Deployed Open WebUI and LibreChat on my Kubernetes Cluster --> The deployment experience is a bit better with LibreChat, as it allows to configure everything with the Helm Chart. In Open WebUI you have to manually add the Credentials for e.g. Anthropic (installed with OUI's "pipelines" ) via the UI.

Would be nice to have CAPH (Cluster API Provider Hetzner) to allow configuring the name of the created LoadBalancer. Currently it does "wait and reconcile until user creates it". But setting a predictable name would allow to easily reuse it for e.g. Ingress controllers.

github.com/syself/clust...