docs(providers): clarify vllm qwen reasoning output

[HwangJohn]

What does this PR do?

This PR documents the expected behavior for Qwen reasoning models served through OpenAI-compatible vLLM endpoints.

Hermes can preserve structured reasoning metadata such as reasoning, reasoning_content, and streamed reasoning deltas, but that metadata is treated as reasoning/thinking trace data. The assistant's user-visible answer should still be emitted in content.

For vLLM/Qwen deployments where --reasoning-parser qwen3 leaves content empty, the docs now point users to two stable options:

• disable the reasoning parser for that deployment, or
• pass a server-supported request option such as chat_template_kwargs.enable_thinking: false through extra_body.

Related Issue

Fixes #38360

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

• website/docs/integrations/providers.md
  • Clarifies how Hermes treats Qwen/vLLM reasoning-parser output.
  • Adds the extra_body.chat_template_kwargs.enable_thinking: false workaround for deployments that separate all generated text into reasoning fields.
• cli-config.yaml.example
  • Adds the nested vLLM/Qwen extra_body shape to the auxiliary-model comment example.

How to Test

1. Read the vLLM provider section in website/docs/integrations/providers.md.
2. Confirm it explains that separated reasoning metadata is not a replacement for visible content.
3. Confirm the custom-provider extra_body section shows the nested chat_template_kwargs.enable_thinking: false shape.

Validation run:

• Local Windows:
  • git diff --check
• DGX Spark Linux:
  • git diff --check
  • cd website && npm ci && npm run build

Results:

• Diff check: passed
• Docusaurus build: passed
• Note: the docs build still reports existing broken link/anchor warnings elsewhere in the site; the build exits successfully and this PR does not add new links.

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: Windows 11 local diff check and DGX Spark Linux docs build

Documentation & Housekeeping

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

For New Skills

N/A

Screenshots / Logs

DGX Spark docs build:

[SUCCESS] Generated static files in "build".
[SUCCESS] Generated static files in "build/zh-Hans".

[teknium1]

Merged via PR #49944 (docs salvage cluster 2). Your commit was cherry-picked onto current main with your authorship preserved in git log — rebase-merged so it lands under your name. Thanks for the contribution!