glm-openrouter-specialist

Caution

OpenRouter adds a routing layer between Claude Code and the GLM backends. Model IDs, auth headers, and provider fallback chains differ from the direct Z.ai API. Sending a direct-API payload to OpenRouter — or vice versa — returns a 400 or silent truncation. This agent handles the translation.

Tip

Auto-delegated by /brewui:glm-design-to-code when OpenRouter is selected as provider. You can also invoke it directly for standalone API requests, cost audits, or model selection decisions.

Quick reference

Field	Value
Name	`glm-openrouter-specialist`
Model	sonnet
Tools	Read, Bash, Glob, Grep, Write, Edit
Triggers	”openrouter”, “open router”, “route glm”, “openrouter api”, “glm via openrouter”, “glm provider”, “openrouter pricing”
Env required	`OPENROUTER_API_KEY`, `BU_PLUGIN_ROOT` (injected by hook)

When to use

Design-to-code via OpenRouter — send a screenshot or image to GLM through the OpenRouter gateway
Model selection — pick the right GLM variant (quality vs. cost vs. free tier) for the task
Cost audit — estimate or review token spend before a batch run
Provider routing — lock to a specific backend (e.g., SiliconFlow) or configure fallback chains
Debugging 4xx/5xx — diagnose auth errors, rate limits, and provider downtime

Examples

"use openrouter to send this screenshot to glm"
"route glm via openrouter with cost optimization"
"which openrouter GLM model should I use for batch design-to-code?"
"openrouter pricing for glm vision tasks"
"glm provider — openrouter api key check"

Flow

Validate env
Checks OPENROUTER_API_KEY is set and valid via /api/v1/auth/key. Reads BU_PLUGIN_ROOT from prompt context (injected by pre-task.mjs). If either is missing — stops with an explicit error before touching any files.
Select model
Matches the task to the optimal GLM variant: vision tasks get z-ai/glm-5v-turbo (quality) or z-ai/glm-4.6v (budget); text-only tasks use z-ai/glm-4.5-air:free at zero cost. Model IDs must carry the z-ai/ prefix — bare names are rejected by OpenRouter.
Build payload
Calls glm-build-request.sh from $BU_PLUGIN_ROOT/skills/glm-design-to-code/scripts/ to construct the JSON payload with image (base64-encoded), prompt, model ID, and max_tokens. Optionally adds provider.order and provider.allow_fallbacks for backend routing control.
Send request
Runs glm-request.sh payload.json response.json openrouter. The script posts to https://openrouter.ai/api/v1/chat/completions with the required Authorization and Content-Type headers, plus optional HTTP-Referer and X-Title for OpenRouter dashboard identification.
Parse response
Checks HTTP status code and finish_reason. Extracts content via jq -r ‘.choices[0].message.content’. For multi-file design-to-code output, delegates extraction to glm-extract.sh which splits on ===FILE:=== / ===END_FILE=== markers.
Report cost
Reads .usage from the response — OpenRouter returns prompt_tokens, completion_tokens, and total_cost. Prints a summary table: model, finish reason, token counts, estimated USD cost. Surfaces any 402/429/5xx errors with a concrete remediation step.

Model matrix & pricing

Model ID	Vision	Input $/1M	Output $/1M	Context	Best for
`z-ai/glm-5v-turbo`	image + video	$1.20	$4.00	202K	Quality design-to-code
`z-ai/glm-4.6v`	image + video	$0.30	$0.90	131K	Cost-effective vision (default)
`z-ai/glm-4.5-air:free`	text only	FREE	FREE	131K	Text generation, no vision

Provider routing snippet:

{
  "provider": {
    "order": ["SiliconFlow"],
    "allow_fallbacks": true,
    "require_parameters": true,
    "data_collection": "deny"
  }
}

Error reference:

HTTP	Meaning	Fix
400	Bad request	Check payload format, verify `z-ai/` model prefix
401	Auth failed	Verify `OPENROUTER_API_KEY`
402	Insufficient credits	Top up OpenRouter account
429	Rate limited	Wait and retry with exponential backoff
502/503	Provider down	Retry or add `provider.order` fallback

🤖

GLM Z.ai Specialist

Direct Z.ai API alternative — no routing layer, lower latency.

🧩

GLM Design to Code

The skill that auto-delegates to this agent when OpenRouter is selected.

🚀

Brewui overview

All brewui skills and agents in one place.

🔗

GitHub source

Agent definition, scripts, and request helpers.

Updating plugins

Use /brewtools:plugin-update to check and update the brewcode plugin suite in one command. See the FAQ for details.