fix(gateway): honor silent busy input modes

[helix4u]

What does this PR do?

Honors silent busy-input modes for CLI and gateway sessions.

display.busy_input_mode was only effectively supporting interrupt and queue in the gateway. Any other value, including the dashboard-exposed block option and user-configured ignore, fell back to interrupt, so busy Discord/Telegram sessions could still post "Interrupting current task..." acknowledgements and interrupt the active run.

This PR makes block and ignore explicit silent-drop modes for normal busy-session input. Dedicated mid-run controls such as /stop, /status, /approve, /deny, /queue, and /steer still work.

Related open PRs:

• fix(gateway): honor queue mode for busy telegram follow-ups #12070 covers queue-mode gateway follow-ups.
• feat(cli): add /busy slash command for busy input mode #11782 adds CLI runtime controls for busy-input mode.
• feat: add busy command to block input during long operations #10303 covers blocking input during compression.

Related Issue

Fixes #

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/run.py: accepts interrupt, queue, block, and ignore from config/env and silently drops normal busy-session input in block/ignore mode.
• cli.py: honors block and ignore for busy Enter behavior instead of falling back to interrupt.
• hermes_cli/web_server.py: exposes all supported busy-input modes in the settings schema.
• website/docs/user-guide/cli.md and website/docs/user-guide/messaging/index.md: document the supported modes.
• Tests cover gateway config loading, adapter-level busy handling, direct running-agent busy handling, and CLI config normalization.

How to Test

1. Set display.busy_input_mode: ignore or display.busy_input_mode: block.
2. Start a gateway session and send a normal follow-up message while the agent is still running.
3. Confirm the follow-up does not interrupt the active run, does not get queued, and does not send a busy acknowledgement.
4. Confirm explicit commands such as /stop and /status still work while the agent is running.

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: Ubuntu 24.04

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

• ./scripts/run_tests.sh tests/gateway/test_busy_session_ack.py tests/gateway/test_restart_drain.py tests/gateway/test_session_race_guard.py tests/cli/test_cli_init.py -n 4 → 70 passed
• ./scripts/run_tests.sh tests/ -n 4 → 135 failed, 15357 passed, 40 skipped, 184 warnings in 304.57s

[teknium1]

Thanks for tracking this down — you correctly identified that block in the dashboard was a lie. After discussion we decided the right fix is the inverse direction: the dashboard option was added speculatively in the original web UI PR with no implementation behind it, and silent-drop behavior is actively bad UX for a messaging bot (users spam when they think the bot's broken, and queue already covers the 'don't interrupt me' case).

Removed block from the dashboard schema in #15355 (merged). The two supported modes remain interrupt (default) and queue.

Appreciate the thorough PR with tests and docs — sorry we're going the other way on this one.