Hey Krill, I'm having trouble getting local Ollama models to show up in the OpenClaw Control UI (version 2026.4.22).
Setup:
Ollama is running locally on 127.0.0.1:11434 serving models like qwen3.6:35b and gpt-oss:20b.
We successfully enabled the plugin (openclaw plugins enable ollama) and checked that discovery: {enabled: true} is active.
The Hangup:
The Control UI only lists venice and says "Unknown provider: ollama".
We realized the global openclaw.json was missing the ollama provider block under models.providers and the ollama:local block under auth.profiles.
The real issue: When we tried to manually edit ~/.openclaw/openclaw.json to add the Ollama provider, the config changes seem to get instantly reverted/overwritten by OpenClaw or the UI, causing the models to disappear again. We resorted to a Python script to force-inject the JSON, which felt like a hacky workaround.
Question:
What is the official, persistent way to configure and register the Ollama provider in 2026.4.x so the auto-discovery works, the UI recognizes it, and the config doesn't keep getting overwritten?