GLM-4.7 (Z.AI) — Quick Start
Send a small test chat request to GLM-4.7 through Tay’s internal adapter and copy the returned JSON.
What it does
- Sends a chat/completions request to Z.AI with model glm-4.7.
- Returns the provider JSON response (OpenAI-like shape).
- Lets you copy the returned JSON for inspection or debugging.
What it does not do
- It does not change Prompt Dictator or any existing tool behavior.
- It does not store your raw prompts in Tay.
- It does not stream responses in this slice.
Step-by-step
- Sign in (the internal adapter route requires an authenticated session).
- Go to /throne.
- In “LLM Providers → GLM-4.7 (Z.AI)”, paste a small test message.
- Press “Send test (GLM)”.
- If it succeeds, copy the JSON result.
Example
Example input
Hello. Reply with a single sentence.
Example output snippet
{
"id": "...",
"object": "chat.completion",
"model": "glm-4.7",
"choices": [ ... ]
}Error meanings
- unauthorized — you’re signed out. Sign in and try again.
- missing_zai_api_key — GLM isn’t configured on this deployment.
- glm_timeout — the Z.AI API did not respond in time.
- glm_upstream_error — Z.AI returned a non-2xx response.
- llm_provider_not_configured — provider disabled or unknown.
Privacy note: No raw prompts are stored by Tay.
When to use it
- When you need a quick sanity check that GLM wiring works.
- When you want to inspect the raw provider response shape.
- When you want a provider call that’s isolated from core tools.
- When you’re debugging timeouts or configuration safely.
When not to use it
- When you need deterministic output (this is a model call).
- When you don’t want data sent to an external provider.
- When the provider isn’t configured or you’re signed out.