fix(zai): complete GLM-5.3 coding metadata

[Ingwannu]

Summary

• set the documented default reasoning effort for zai/glm-5.3 and zai/glm-5.3[1m] to max
• set the documented per-model maximum output to 128,000 tokens
• verify registry seeding, key-login derivation, routing, and catalog default metadata

Scope

This change is limited to the Z.AI Coding Plan route. It does not infer GLM-5.3 availability or metadata for unrelated pay-as-you-go or aggregator endpoints.

Validation

• bun test tests/provider-registry-parity.test.ts (37 passed)
• bun run typecheck
• bun run privacy:scan
• git diff --check

Fixes #1734

Summary by CodeRabbit

• New Features

  • GLM-5.3 models, including the [1m] variant, now default to maximum reasoning effort.
  • Increased the maximum generated output to 131,072 tokens (128K).

• Bug Fixes

  • Provider routing and model catalog entries now consistently reflect the updated reasoning and output limits.

[github-actions]

✅ Deterministic PR hygiene checks passed.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 6652e7a7-d21d-4f58-8ecc-06a84421331d

📥 Commits

Reviewing files that changed from the base of the PR and between 5f55cd8 and ad89bc0.

📒 Files selected for processing (2)

• src/providers/registry.ts
• tests/provider-registry-parity.test.ts

---

📝 Walkthrough

Walkthrough

The Z.AI registry assigns max reasoning effort and a 131,072-token output limit to GLM-5.3 variants. Provider parity tests verify propagation through provider configuration, routing, key-login derivation, and catalog generation.

Changes

Z.AI GLM-5.3 metadata

Layer / File(s)	Summary
Registry defaults and metadata propagation src/providers/registry.ts, tests/provider-registry-parity.test.ts	The registry defines max as the default reasoning effort and 131,072 as the maximum output-token limit for glm-5.3 and glm-5.3[1m]. Tests verify propagation through provider configuration, routing, key-login derivation, bracket suffix stripping, and generated catalog entries.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• lidge-jun/opencodex#407: Updates overlapping GLM-5.3 registry metadata and parity tests.
• lidge-jun/opencodex#536: Modifies the same provider registry and parity-test areas for Zhipu/GLM metadata.
• lidge-jun/opencodex#853: Updates reasoning-effort metadata and its propagation into catalog entries.

Suggested labels: provider

Suggested reviewers: lidge-jun, wibias

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly identifies the Z.AI GLM-5.3 Coding Plan metadata fix, which is the primary change.
Linked Issues check	✅ Passed	The registry and parity tests add max reasoning and 128,000-token output metadata for both GLM-5.3 Coding Plan aliases and verify propagation.
Out of Scope Changes check	✅ Passed	The changes are limited to Z.AI registry metadata and focused parity tests required by issue #1734.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch ingw/fix-zai-glm53-metadata-1734

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}