fix(agent): route Anthropic fast mode via extra_body

[marcelruhf]

What does this PR do?

Fixes the native Anthropic fast-mode request shape introduced in #7037.

/fast fast stores {"speed": "fast"} in request_overrides, and the Anthropic adapter was forwarding that as a top-level speed= kwarg. On the stable Anthropic Python SDK, client.messages.create() and client.messages.stream() do not accept top-level speed; both surfaces accept undocumented body params through extra_body. The old shape fails locally with TypeError before a request is sent.

This PR keeps the fix at the adapter boundary:

• route fast mode as extra_body["speed"] = "fast"
• preserve the fast-mode beta header
• keep third-party Anthropic-compatible endpoints excluded from fast mode

extra_body is the right fix because build_anthropic_kwargs() already owns Anthropic-specific request shaping, and the SDK exposes extra_body on both the current release (0.93.0) and the repo's minimum supported floor (0.39.0).

Related Issue

Regression from #7037.

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

• agent/anthropic_adapter.py: move fast-mode speed from a top-level kwarg to extra_body, with SDK-compatibility notes in the docstring/comments
• tests/agent/test_anthropic_adapter.py: add a real SDK-boundary regression test using anthropic + httpx.MockTransport to verify "speed": "fast" is serialized into the JSON body
• tests/cli/test_fast_command.py: update fast-mode assertions to expect extra_body, and keep the third-party endpoint exclusion coverage

How to Test

1. Activate the project environment:
   source /home/mruhf/.hermes/hermes-agent/venv/bin/activate
2. Run the focused regression suites:
   python -m pytest tests/cli/test_fast_command.py -q
python -m pytest tests/agent/test_anthropic_adapter.py -q
python -m pytest tests/cli/test_fast_command.py tests/agent/test_anthropic_adapter.py -q
python -m pytest tests/agent/test_minimax_provider.py -q
python -m pytest tests/gateway/test_fast_command.py -q
3. Verify the old and new request shapes against the current Anthropic SDK:
   • old shape: client.messages.create(..., speed="fast") and client.messages.stream(..., speed="fast") raise TypeError
   • patched shape: build_anthropic_kwargs(..., fast_mode=True) succeeds and sends "speed": "fast" in the JSON body

Additional note:

• python -m pytest tests/ -q currently fails on this worktree with unrelated pre-existing failures outside this PR's scope; the focused suites above are green.

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: Linux 6.8.0-106-generic

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

Manual SDK verification on anthropic==0.93.0:

Messages.create() got an unexpected keyword argument 'speed'
Messages.stream() got an unexpected keyword argument 'speed'

Patched adapter verification:

response_text ok
body_speed fast
header_has_fast_beta True
top_level_speed_in_kwargs False

[teknium1]

Thanks for the careful analysis and the SDK-boundary regression test, @marcelruhf! This fix was independently implemented on main before your PR could be reviewed.

This is an automated hermes-sweeper review.

• The identical fix (kwargs.setdefault("extra_body", {})["speed"] = "fast") landed in commit cdd44817f on Apr 13 2026 — one day after this PR was opened.
• It shipped in release v2026.4.16.
• The fix lives at agent/anthropic_adapter.py line 1704, exactly as proposed here.
• tests/cli/test_fast_command.py was also updated in that commit with extra_body assertions covering the same regression.

The tests/agent/test_anthropic_adapter.py SDK-boundary test using httpx.MockTransport is the one novel addition here that didn't make it to main — if you'd like to contribute that test in isolation it would still be welcome as a follow-up.