* feat(ai): sync gateway providers and model discovery
Add provider account types and protocol-aware model discovery, and replace predefined models with editable model pools.
Align provider icons with AI Gateway, preserve international Coding Plan restrictions, and filter account counts consistently with selectable accounts.
* fix(ai): preserve Ollama base paths during model discovery
* fix(ai): restore custom DashScope image protocol support
llmman (https://github.com/llmmanorg/llmman) is a local model runner
serving Ollama- and OpenAI-compatible routes on 127.0.0.1:17434.
Register it in the agent provider catalog next to Ollama and extend
every Ollama special case (no API key, verification skipped, OpenClaw
placeholder key, manual initial model) to cover it as well.
Add the latest MiniMax-M3 flagship model to the MiniMax provider model
list and set it as the default. M3 offers a 1M context window, 128K max
output, reasoning, and image input. Drop the older M2.5 entries while
keeping M2.7 as an alternative, and update the account verification probe
to use M3.
Co-authored-by: octo-patch <266937838+octo-patch@users.noreply.github.com>