fix: relabel GLM provider display to Z.ai

[lsm]

GLM is the model family (GLM-5, GLM-5.2…), not the provider/company (Zhipu AI, internationally branded Z.ai). The provider was labelled "GLM" next to models named "GLM-5" — confusing and redundant.

Change

Display-only rename of every user-facing provider label from GLM → Z.ai. Model names (GLM-5, GLM-5.2, …) are unchanged — GLM is the model name.

Surface	Before	After
Model-picker group header	GLM	Z.ai
Add-provider card title	GLM	Z.ai
Provider displayName	GLM (智谱AI)	Z.ai
Credential-probe errors	GLM API key rejected	Z.ai API key rejected

Unchanged (by design — no migration, existing configs keep working)

Provider id glm, model ids/aliases/family, registry routing, settings.json keys, env vars (GLM_API_KEY / ZHIPU_API_KEY), stored credentials, brand color #7DD3FC, and the ProviderLogo Z-mark.

Test plan

• bun run check (lint + typecheck + knip + guards) passes
• daemon glm-provider unit + related (provider-service, credential-discovery) pass
• web useModelSwitcher, provider-brand, AddProviderModal, ProvidersSettings pass
• grep confirms no user-facing GLM provider labels remain; remaining "GLM" strings are model names, env vars, or routing keys

Closes #770

[lsm]

🤖 Review by glm-5.1 (Z.ai)

Model: glm-5.1 | Client: NeoKai | Provider: Z.ai

Recommendation: REQUEST_CHANGES — posting as COMMENT because the PR is authored under the same account (lsm), so GitHub rejects REQUEST_CHANGES/APPROVE on my own PR.

What is correct

The display-only relabel is clean and accurate in every file it touches: provider displayName (glm-provider.ts), credential-probe providerName, buildSdkConfig missing-key error, PROVIDER_LABELS (useModelSwitcher.ts), AddProviderModal preset, credential-discovery.ts displayName, and KNOWN_PROVIDER_LABELS (provider-brand.ts). The doc comment in shared/types.ts is updated too.

Unchanged by design and verified intact: provider id glm, model ids/aliases/family, registry routing, settings.json keys, env vars (GLM_API_KEY/ZHIPU_API_KEY), stored credentials, brand color #7DD3FC. No migration — existing sessions/configs keep working.

I specifically verified the provider-brand.ts change is safe: KNOWN_PROVIDER_LABELS only feeds TRAILING_PROVIDER_TAG (strips trailing (Z.ai) suffixes — no GLM model carries one), and REDUNDANT_BRAND_PREFIXES has no glm entry, so model rows still render as "GLM-5", "GLM-5.2", etc. ✓ Acceptance criterion met on this axis.

Tests pass locally with assertions correctly updated: daemon glm-provider unit 42/42; web useModelSwitcher + provider-brand + AddProviderModal 102/102.

One missed user-facing provider label — P2 (blocks)

packages/daemon/src/lib/agent/query-runner.ts:524

const authError = new Error(
  'No authentication configured. Please set up API key for Anthropic or GLM.'
);

Here "GLM" is listed as a peer of "Anthropic" — i.e. as the provider/company — in a live, user-facing auth error (fires when no provider auth is configured; surfaced via errorManager.handleError as AUTHENTICATION). This is exactly the provider/company sense the rename targets, and it directly violates acceptance criterion #4 ("a grep for user-facing GLM provider labels confirms none remain that refer to the provider/company"). The completeness grep missed it because query-runner.ts was not in the Key files list.

Fix: Anthropic or GLM → Anthropic or Z.ai. No test asserts this exact string, so it is a clean one-liner with no test coordination needed.

Considered and intentionally not flagged

glm-provider.ts:193 — getAuthStatus returns Set GLM_API_KEY or ZHIPU_API_KEY to enable GLM models. The env-var names are correct to keep, and "GLM models" refers to the model family (GLM-5, GLM-5.2…), which is accurate. Not a defect.

Fix the single line in query-runner.ts:524 and this is ready to re-review / hand to QA.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1b81dc7dc3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".