Set ZhiPu LLM Service Connector (not token plan) 🐑

This is the standard, pay-as-you-go connector for ZhiPu (智谱), the lab that spun out of Tsinghua's Knowledge Engineering Group in 2019 and now publishes internationally as Z.ai. Worth knowing going in: this isn't a small or unproven company you're trusting your API key to. ZhiPu IPO'd in Hong Kong in January 2026 as China's first publicly listed frontier AI company, and its GLM model family has shipped under MIT license (no revenue cap, no content restrictions) since July 2025 - the current flagship, GLM-5.2, is a 744-billion-parameter mixture-of-experts model with a million-token context window.

The node's own name flags the thing to get right first: "not token plan." ZhiPu sells two separate products through this pack - a plain pay-per-token Open Platform (what this node connects to, reading a zhipu entry in your config) and a flat-fee Coding/Token Plan subscription (a different connector, SetZhiPuCodeLLMServiceConnector, reading a separate zhipu_code entry). Same models under the hood, different billing, different key issued from a different signup page on bigmodel.cn. If you signed up for the coding subscription and plug that key into this node, expect it to fail - they're not interchangeable.

What it actually does. Nothing on its own, same as every Set*LLMServiceConnector in this pack - it resolves your key and model choice into an LLMServiceConnector object that a downstream node consumes. Wire it into the pack's Translator or a prompt-generator node to see any output at all.

Inputs that matter. api_token is your ZhiPu Open Platform key from bigmodel.cn. model_select defaults to glm-5.1, with eight other options: the current line (glm-5.2, glm-5-turbo, glm-5) and the previous generation (glm-4.7, glm-4.6, glm-4.5, and the lighter glm-4.5-air). For general prompt rewriting or translation work - which is mostly what feeds off this connector in the pack - the mid-tier models are usually plenty; save the flagship for anything that actually needs the extra reasoning depth. If ZhiPu ships something after this dropdown was last updated, Custom plus custom_model gets you there.

The config-file gotcha. prefer_local_config defaults to true: the node reads mie_llm_keys.json (copied from mie_llm_keys.json.example, kept next to the README, deliberately excluded from git) and uses the zhipu entry there over whatever's typed into the node's own api_token. This is the most common source of confusion across every connector in this pack - a freshly typed key that gets silently overridden by a stale one in the file. Fix the JSON, or flip prefer_local_config to false if you want the node's own field to always win. config_key lets you read a different entry if you keep multiple ZhiPu accounts.

Install. Search "ComfyUI_MieNodes" in ComfyUI Manager, or cd ComfyUI/custom_nodes && git clone https://github.com/MieMieeeee/ComfyUI-MieNodes and restart. No downloads needed - this is API plumbing, nothing runs locally.

Troubleshooting. An auth error on a key that looks correct almost always traces to one of two things: either the config-file precedence above, or you're accidentally using a Coding Plan key here instead of an Open Platform one - double-check which product you actually signed up for on bigmodel.cn. Model 404s mean ZhiPu has retired or renamed something since this dropdown was last refreshed; custom_model with the current ID from ZhiPu's own docs is the fix.