dsh-zai-coding-models

Zhipu ships GLM-5.3 on the coding endpoint and your token plan is already active, yet the pi-ai catalog bundled inside DSH lags by one or two release cycles — the model picker still shows only the previous generation. That gap means paying for the newest model but not being able to reach it: you wait for a DSH upgrade, restart and re-verify, or hand-edit settings.yaml line by line.

This plugin packages DSH provider-profile patching into a single install step. The six existing zai-coding-cn models stay untouched while GLM-5.3 enters the picker with 1M context, 128K max output, and four thinking levels (Low / Medium / High / Max). It also sets the compat flags the endpoint requires — thinkingFormat and supportsReasoningEffort — so the reasoning-effort parameter is no longer silently dropped. Your API key, endpoint, and every other provider configuration remain exactly where you left them.

Built for developers who already hold a Zhipu coding plan, have the key configured through the DSH Models page, but cannot find GLM-5.3 in the picker. Once the pi-ai catalog catches up in a future DSH release, simply remove the plugin and fall back to the stock directory with zero residual configuration.