feat(opencode-go): support GLM-5 reasoning controls (#49279)

[izumi0uu]

What does this PR do?

Adds GLM-5 reasoning support to the opencode-go profile so Hermes stops silently dropping /reasoning for GLM-5 model IDs routed through plugins/model-providers/opencode-zen/__init__.py.

Before this change, OpenCodeGoProfile.build_api_kwargs_extras() only applied reasoning controls to Kimi K2 and DeepSeek thinking models. glm-5* IDs fell through the gate entirely, so /reasoning high, /reasoning off, and related settings had no effect on GLM-5 requests on this provider path.

This PR adds a GLM-5 model gate and reuses the existing DeepSeek-style mapping already used on OpenCode Go:

• recognized efforts forward as top-level reasoning_effort
• disabled reasoning emits extra_body.thinking={"type":"disabled"}
• unrecognized/no explicit effort falls back to extra_body.thinking={"type":"enabled"}

Related Issue

Fixes #49279

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• Added _is_glm_thinking_model() to plugins/model-providers/opencode-zen/__init__.py
• Routed GLM-5 IDs through the existing OpenCode Go reasoning-control branch
• Added GLM-specific unit coverage in tests/plugins/model_providers/test_opencode_go_profile.py
• Extended model-gating tests to keep glm-4.5 and glm-4.5-flash excluded
• Added transport integration coverage to verify reasoning_effort reaches final kwargs for GLM-5

How to Test

1. On main, call OpenCodeGoProfile.build_api_kwargs_extras(..., model="glm-5.1") with a reasoning config and observe that it returns no GLM reasoning controls.
2. On this branch, repeat the same call and verify that GLM-5 models now emit the same reasoning fields as the existing DeepSeek branch.
3. Run:
   • ./.venv/bin/python -m pytest tests/plugins/model_providers/test_opencode_go_profile.py -q
   • git diff --check

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 15.7.7

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Screenshots / Logs

Targeted validation run:

$ ./.venv/bin/python -m pytest tests/plugins/model_providers/test_opencode_go_profile.py -q
28 passed, 3 warnings in 3.70s

$ git diff --check
(no output)

[teknium1]

Thanks for isolating the missing provider-profile path and adding transport-level coverage.

Problems

• Current main now has a GLM-5.2-specific OpenCode Go implementation at plugins/model-providers/opencode-zen/__init__.py:64-77; it emits only native top-level reasoning_effort and intentionally leaves disabled/unset requests at server defaults.
• Current regression coverage explicitly requires glm-5.1 and glm-5 to emit no OpenCode Go reasoning controls (tests/plugins/model_providers/test_opencode_go_profile.py:185-201). The proposed startswith("glm-5") gate changes that contract by applying the DeepSeek wire shape.
• The linked issue's only comment says the GLM-5.x relay contract was not verified, including whether it accepts top-level reasoning_effort or requires another format.

Suggested changes

• Verify the OpenCode Go GLM-5.1 relay contract, then add a narrowly scoped GLM-5.1 branch and tests for enabled, disabled, and unset behavior without changing the existing GLM-5.2 contract.

Automated hermes-sweeper review.

[teknium1]

This broad glm-5* gate applies DeepSeek's thinking/reasoning_effort contract to GLM-5.1, but the linked issue says the relay contract is unverified. Current main instead scopes native controls to GLM-5.2 and explicitly tests glm-5.1 as non-target; please verify and encode the exact GLM-5.1 wire contract before widening this gate.

[izumi0uu]

Thanks for the review. I verified the OpenCode Go contract instead of carrying over the DeepSeek mapping. The official metadata marks GLM-5.1 as reasoning-capable but exposes no configurable reasoning_options; OpenCode maintainers also confirmed that excluding GLM-5/5.1 reasoning variants is intentional.

GLM-5.2 is the supported exception, and its high/max reasoning_effort handling has already landed on main in #58884 (a6079dd). Narrowing this PR to GLM-5.2 would therefore duplicate the existing implementation, while retaining GLM-5.1 support would send an unsupported payload. I’m withdrawing this PR.