fix(gateway): reject multi-line /model with clear error

[dockerpaula]

What does this PR do?

Fixes #22716 — a multi-line /model command (e.g. Shift+Enter in Element/Matrix produces /model openai/gpt-5.5\nFollow-up question in a single send) is currently rejected with the misleading error Model names cannot contain spaces, and the user's follow-up text is silently swallowed. The error message is wrong in two ways: there are no spaces in the model name, and the real problem is that \n was forwarded into the model-name parser at all.

This PR detects the multi-line form early in _handle_model_command and returns a clear, actionable message that quotes the intended single-line command back to the user, e.g.:

`/model` is a single-line command. Send `/model openai/gpt-5.5` first, then send your follow-up message separately.

The fix is intentionally surgical — no behavior change for legitimate single-line /model invocations, and no change to other commands that do accept multi-line text (/queue, /steer, /agent, etc.) because the new helper only short-circuits when called from the /model handler.

Related Issue

Fixes #22716

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

• gateway/platforms/base.py — added MessageEvent.get_command_remainder() helper. Returns the text that follows the first newline of a command message (e.g. /model X\nFoo → "Foo"), with strip semantics; returns "" for non-commands, single-line commands, command-only messages, or messages whose remainder is whitespace-only. Sits next to the existing get_command_args()/is_command() helpers.
• gateway/run.py — added an early multi-line detection block in _handle_model_command() that uses get_command_remainder() to short-circuit before reaching the model-name parser. Handles two distinct multi-line shapes:
  • /model X\nFoo — get_command_args() returns "X\nFoo", so the suggestion is `/model X`.
  • /model\nFoo — get_command_args() consumes the \n as the arg separator and returns "Foo", so the suggestion falls back to `/model <name>`.
• tests/gateway/test_platform_base.py — added TestMessageEventGetCommandRemainder (8 tests) covering: single-line no-remainder, command-only no-remainder, multi-line returns remainder, remainder is stripped, inner newlines in remainder preserved, blank-only remainder returns "", non-command returns "", and a non-regression check that get_command_args() still returns the full multi-line tail for commands like /queue line1\nline2\nline3 that legitimately accept multi-line text.
• tests/gateway/test_model_command_multiline.py (new file, 4 tests) — async regression tests for _handle_model_command() covering: the canonical /model X\nFoo case, multi-line with flags (/model X --global\nFoo), benign trailing-newline (/model X\n with no follow-up — should NOT trigger the multi-line error), and the empty-first-line fallback (/model\nFoo — should suggest the generic /model <name> form).
• scripts/release.py — adds dockerpaula@gmail.com → dockerpaula to AUTHOR_MAP so the contributor-attribution check resolves the email when generating release notes.

How to Test

Reproduction (before the fix):

# In any platform that delivers a single message containing \n (Element/Matrix
# Shift+Enter, Slack composer multi-line, etc.):
/model openai/gpt-5.5
What is 2+2?

Before — the gateway replies: Model names cannot contain spaces (and the question is dropped).
After — the gateway replies: `/model` is a single-line command. Send `/model openai/gpt-5.5` first, then send your follow-up message separately.

Automated:

# 1. New regression tests for both the helper and the handler — 102 passed, 2 skipped
scripts/run_tests.sh tests/gateway/test_platform_base.py tests/gateway/test_model_command_multiline.py -q

# 2. All existing /model command tests — 113 passed, 2 skipped — confirms no regression
scripts/run_tests.sh tests/gateway/test_model_command_*.py tests/gateway/test_model_switch_*.py tests/gateway/test_discord_model_picker.py -q

# 3. Lint + Windows-footguns
ruff check gateway/platforms/base.py gateway/run.py tests/gateway/test_platform_base.py tests/gateway/test_model_command_multiline.py scripts/release.py
python scripts/check-windows-footguns.py --all

All green locally.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(gateway): …, chore(release): …)
• I searched for existing PRs to make sure this isn't a duplicate — none reference Matrix gateway: slash command + free-text question in a single multi-line message is rejected, blocking inline orchestration patterns #22716
• My PR contains only changes related to this fix
• I've run targeted pytest and all tests pass
• I've added tests for my changes (12 new tests: 8 helper + 4 handler)
• I've tested on my platform: Ubuntu 24.04

Documentation & Housekeeping

• N/A — user-visible behavior change is the error message itself; no docs page documented the previous wrong message
• N/A — no config keys added/changed
• N/A — no architecture or workflow change
• Cross-platform: pure-Python string handling (str.partition, str.split); no platform-specific calls
• N/A — no tool descriptions/schemas changed

Screenshots / Logs

$ pytest tests/gateway/test_platform_base.py tests/gateway/test_model_command_multiline.py -q
........................................................................ [ 69%]
.........................s.s....                                         [100%]
102 passed, 2 skipped in 3.09s

[teknium1]

Thanks for the focused regression coverage. The underlying bug still exists on current main: gateway/platforms/base.py:1804 retains the multi-line argument tail, and the current handler at gateway/slash_commands.py:1420 passes it into flag parsing and model validation.

Problems

• tests/gateway/test_model_command_multiline.py:90 only proves that a trailing newline does not return the new guidance text. It also passes for an unrelated model-switch error, so it does not establish the documented “switch must proceed” behavior.

Suggested changes

• During salvage, place the early guard in the current handler at gateway/slash_commands.py:1420; 619bd782 moved _handle_model_command out of gateway/run.py.
• Make the trailing-newline test control or mock the normal switch pipeline and assert that it receives the first-line model name.

This is an automated hermes-sweeper review.

[teknium1]

This assertion accepts any non-guidance failure, including a validation or provider error, so it does not prove the stated contract that a trailing newline reaches the normal successful switch path. Please control the switch dependency and assert the first-line model argument is used.