fix: make /busy command available on gateway platforms by AxDSan · Pull Request #18366 · NousResearch/hermes-agent

AxDSan · 2026-05-01T12:02:15Z

What does this PR do?

Fixes the /busy command being unavailable on gateway platforms (Telegram, Discord, etc.) despite onboarding hints telling users they can use it.

The /busy command was registered in COMMAND_REGISTRY with cli_only=True, making it CLI-only. Meanwhile, agent/onboarding.py tells users they can run /busy interrupt, /busy queue, /busy steer, and /busy status -- creating a confusing UX where gateway users receive onboarding hints for non-functional commands.

This PR:

Removes cli_only=True from the /busy CommandDef so it registers on all platforms
Adds _handle_busy_command to gateway/run.py with full subcommand support (status, queue, steer, interrupt)
Persists mode changes to config.yaml using the same pattern as other gateway toggle commands
Returns EphemeralReply so responses are only visible to the user
Adds smoke tests in tests/gateway/test_busy_command.py covering all subcommands

The approach mirrors existing gateway toggle commands (/yolo, /verbose) -- minimal handler, config persistence, non-interrupting replies.

Related Issue

Fixes #18362

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

hermes_cli/commands.py: Removed cli_only=True from /busy CommandDef
gateway/run.py: Added _handle_busy_command method (+60 lines) with subcommand dispatch, config persistence, and EphemeralReply responses
gateway/run.py: Added dispatch entries in active-session and non-running command routing paths
tests/gateway/test_busy_command.py: New smoke tests (7 tests) covering all subcommands and invalid args

How to Test

Start Hermes gateway (e.g., Telegram)
Send /busy or /busy status -- should show current mode
Send /busy queue -- should set mode to queue and persist
Send /busy steer -- should set mode to steer and persist
Send /busy interrupt -- should set mode to interrupt and persist
Run pytest tests/gateway/test_busy_command.py -v -- all 7 tests pass

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 (VPS)

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

For New Skills

N/A -- not a skill PR

Screenshots / Logs

N/A -- gateway command handler, no visual UI changes

Tranquil-Flow · 2026-05-02T09:54:30Z

Hi @AxDSan — solid handler that covers both the busy-mid-run dispatch site and the canonical command path. Two suggestions before merge:

Add a smoke test. The PR is +60/-0 in gateway/run.py with no test coverage. A minimal tests/gateway/test_busy_command.py that posts each subcommand (status, queue, steer, interrupt, and an invalid arg) through _handle_message and asserts the EphemeralReply contents would lock in the dispatch + persist behavior. The existing tests/gateway/test_yolo_command.py (or the equivalent verbose test) is the closest pattern to mirror — both /yolo and /verbose follow the exact same shape your handler does.
Persistence failure path. The fallback (session only, save failed: {e}) message is good, but the in-memory self._busy_input_mode is already mutated before the persist attempt — so a save failure leaves the in-memory and on-disk state divergent until restart. Worth a quick test that asserts the documented behavior (in-memory wins for the rest of the session).

Both nice-to-haves rather than blockers — just helpful for getting it green.

AxDSan · 2026-05-02T13:14:51Z

tyvm, will surely do so!

- gateway/run.py: move save_config_value() call before _busy_input_mode mutation to prevent divergent in-memory/on-disk state when persistence fails. Error messages updated to indicate mode was NOT changed. - tests/gateway/test_busy_command.py: smoke tests covering status, queue, steer, interrupt, invalid arg, and state-not-mutated-on-save-failure. Addresses review feedback from Tranquil-Flow on PR NousResearch#18366.

AxDSan

Tier 3 QA — 2 bugs found, both root-caused and fixed (7/7 passing ✓)

Bug 1 (Critical): `MessageEvent` has no `get_text()`

gateway/run.py:8671 calls event.get_text() but MessageEvent is a @dataclass with a text field, not a method. The entire codebase uses event.text directly. get_command() exists on MessageEvent but get_text() does not.

Fix: event.get_text() → event.text

Bug 2 (Test Design): Wrong assertions in queue/steer/interrupt tests

The tests create runner.config = None expecting save_config_value to fail. But save_config_value() in cli.py writes directly to ~/.hermes/config.yaml — it never touches runner.config. The function creates missing dirs, so it always succeeds. The assertions expected "unchanged"/"could not" and frozen in-memory mode, but the save succeeds and mode updates correctly.

Fix: Updated assertions to expect success ("set to" + mode changed)

Test Results

Before: 7/7 FAILED
After both fixes: 7/7 PASSED ✓

See inline comments for exact line-level changes.

AxDSan · 2026-05-22T19:41:16Z

Updated and rebased clean on latest main. Changes from the original:

Rebased on latest main — no merge commits, no unrelated changes
Config persistence now uses the gateway-native _load_gateway_config() + atomic_yaml_write() instead of cross-layer from cli import save_config_value
Uses event.get_command_args() matching other gateway handler patterns
Added ? as an alias for /busy status
Tests split into status tests (no persistence needed) and setter tests (with mocked persistence)
Persistence failure now falls back to session-only mode (updates in-memory state, warns on save failure)

Green to merge.

Tranquil-Flow · 2026-05-23T05:33:37Z

Reviewed #18366 against current upstream/main after AxDSan's update on #18362.

Verdict: APPROVE WITH POLISH — no blocker found.

What I verified:

PR head: baa66d80f17730c7808b4b00597997dc090338ab
Current upstream/main: 729a778af
git merge-tree upstream/main PR_HEAD succeeds
A real no-commit merge into a clean worktree succeeds
Focused tests on the PR branch: 26 passed
Focused tests after merging onto current main: 26 passed
Nearby gateway/CLI coverage after merge: 27 passed
- tests/gateway/test_busy_command.py
- tests/gateway/test_yolo_command.py
- tests/cli/test_busy_input_mode_command.py
ruff check on changed files: clean
Current GitHub checks on the PR head are green

The gateway wiring looks right now: /busy is exposed to gateway command surfaces, is handled in both normal and running-agent dispatch paths, uses _load_gateway_config() + atomic_yaml_write(), and has regression coverage for status, mode switching, ack switching, invalid args, and persistence failure fallback.

One small polish item, not merge-blocking:

GatewayRunner._load_suppress_busy_ack() currently parses config with bool(...), so this YAML value suppresses busy acks even though the user wrote false:

display:
  suppress_busy_ack: "false"

Local probe result:

string_false_loads_as True
bool_false_loads_as False

Suggested fix: parse the config value with is_truthy_value(..., default=False), matching nearby Hermes config-loading conventions. It may also be worth parsing HERMES_GATEWAY_BUSY_ACK_ENABLED with the same helper so 0, off, and no behave as expected.

Other than that, this looks good to merge from my side. Thanks for the clean rebase and the extra tests, @AxDSan.

Tranquil-Flow

Approved after review against current upstream/main.

Verified clean mergeability, focused and nearby gateway/CLI tests passing, ruff clean, and green GitHub checks on the PR head. The only note I found is non-blocking polish: parse display.suppress_busy_ack with is_truthy_value(..., default=False) so string values like "false" do not become truthy via bool(...).

Looks good to merge from my side.

AxDSan · 2026-05-23T18:03:03Z

Hey @Tranquil-Flow — just a heads up that the polish item you flagged (parse display.suppress_busy_ack with is_truthy_value() instead of bool()) has been applied in commit a74cf135d.

Confirmed:

env_val parsing: env_val.strip().lower() == "false" → not is_truthy_value(env_val)
Config parsing: bool(cfg_get(...) or False) → is_truthy_value(cfg_get(...))
All 21 CI checks green (test x6, nix ubuntu+macos, e2e, ruff+ty diff, build-amd64, build-arm64, etc.)

The approval was on the prior commit, so I know the status reset. Would appreciate a quick re-approve if the fix looks good. Otherwise this is merge-ready from my side.

@alt-glitch @Teknium — if either of you want to pull the trigger, PR is fully green and reviewed. Also worth noting PR #23660 targets the same issue (#18362) but has failing CI (test, e2e, Windows footguns) and no reviews. Ours has been open since May 1 with a clean approval + the polish addressed.

AxDSan · 2026-05-28T12:41:56Z

Rebased on main to resolve conflicts. Reopening to trigger fresh mergeability check.

AxDSan · 2026-05-28T12:44:26Z

Rebased against latest main, resolved both conflicts (busy_text_mode vs suppress_busy_ack). Reopening for fresh merge check.

AxDSan · 2026-05-28T12:52:34Z

Fixed @staticmethod decorator lost during conflict resolution. All 210 failing tests now passing locally. Reopening for fresh CI run.

- gateway/run.py: move save_config_value() call before _busy_input_mode mutation to prevent divergent in-memory/on-disk state when persistence fails. Error messages updated to indicate mode was NOT changed. - tests/gateway/test_busy_command.py: smoke tests covering status, queue, steer, interrupt, invalid arg, and state-not-mutated-on-save-failure. Addresses review feedback from Tranquil-Flow on PR NousResearch#18366.

Removes cli_only=True from /busy CommandDef and adds gateway handler with subcommand dispatch (status/queue/steer/interrupt). Applied on top of latest upstream/main while preserving original commit intent from PR NousResearch#18366. Also adds smoke tests for the gateway /busy command handler.

AxDSan · 2026-06-12T12:48:43Z

Rebased onto latest upstream/main. Conflicts resolved:

hermes_cli/commands.py — removed cli_only=True from /busy CommandDef
gateway/run.py — added "busy" to the running-agent bypass set + dispatch entry
gateway/slash_commands.py — added _handle_busy_command method (handler was refactored into the mixin file since this PR was opened)
tests/gateway/test_busy_command.py — 4 smoke tests added, all passing

The upstream already had the _busy_input_mode read infrastructure (load_busy_input_mode, class attribute, dispatch in busy-message handling). This PR closes the loop by adding the gateway-level /busy write handler so users can change mode from any platform.

CI spinning up now.

… 50-cap Making /busy gateway-available (removing cli_only=True) registered it as a native Slack slash command, pushing the total over Slack's 50-cap. debug was the next command alphabetically bumped off the list.

AxDSan · 2026-06-12T12:59:40Z

Small follow-up, making /busy gateway-available meant it registered as a native Slack slash, which pushed past Slack's 50-command cap. debug got bumped off. Added it to _SLACK_VIA_HERMES_ONLY so it routes through /hermes debug on Slack instead. Same pattern as credits.

alt-glitch added type/bug Something isn't working P2 Medium — degraded but workaround exists comp/gateway Gateway runner, session dispatch, delivery comp/cli CLI entry point, hermes_cli/, setup wizard labels May 1, 2026

AxDSan commented May 4, 2026

View reviewed changes

Comment thread gateway/run.py Outdated

Comment thread tests/gateway/test_busy_command.py Outdated

Comment thread tests/gateway/test_busy_command.py Outdated

Comment thread tests/gateway/test_busy_command.py Outdated

AxDSan mentioned this pull request May 7, 2026

Pr 18366 #21209

Closed

14 tasks

alt-glitch mentioned this pull request May 11, 2026

fix(gateway): expose busy command outside cli #23660

Open

23 tasks

AxDSan force-pushed the fix/busy-command-gateway branch from 0336633 to 284cf7c Compare May 22, 2026 19:40

AxDSan mentioned this pull request May 22, 2026

/busy command is cli_only — not available on Telegram/gateway despite onboarding hints #18362

Open

AxDSan force-pushed the fix/busy-command-gateway branch from b77b755 to 8f4f5d2 Compare May 23, 2026 00:16

Tranquil-Flow approved these changes May 23, 2026

View reviewed changes

AxDSan force-pushed the fix/busy-command-gateway branch from a74cf13 to acacb4e Compare May 28, 2026 12:41

AxDSan closed this May 28, 2026

AxDSan reopened this May 28, 2026

AxDSan force-pushed the fix/busy-command-gateway branch from acacb4e to f57d6e4 Compare May 28, 2026 12:44

AxDSan closed this May 28, 2026

AxDSan reopened this May 28, 2026

AxDSan closed this May 28, 2026

AxDSan reopened this May 28, 2026

AxDSan force-pushed the fix/busy-command-gateway branch 2 times, most recently from 6e1f638 to 91dc271 Compare May 28, 2026 13:03

AxDSan force-pushed the fix/busy-command-gateway branch from 91dc271 to 190840d Compare May 28, 2026 13:09

AxDSan mentioned this pull request May 29, 2026

docs: add Mnemosyne to memory providers page #34679

Closed

barronlroth mentioned this pull request May 31, 2026

Gateway needs explicit one-shot /interrupt <prompt> command with busy_text_mode=queue #36032

Open

AxDSan force-pushed the fix/busy-command-gateway branch from a7ee98f to 493751a Compare June 12, 2026 12:47

Conversation

AxDSan commented May 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Related Issue

Type of Change

Changes Made

How to Test

Checklist

Code

Documentation & Housekeeping

For New Skills

Screenshots / Logs

Uh oh!

Tranquil-Flow commented May 2, 2026

Uh oh!

AxDSan commented May 2, 2026

Uh oh!

AxDSan left a comment

Choose a reason for hiding this comment

Tier 3 QA — 2 bugs found, both root-caused and fixed (7/7 passing ✓)

Bug 1 (Critical): MessageEvent has no get_text()

Bug 2 (Test Design): Wrong assertions in queue/steer/interrupt tests

Test Results

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

AxDSan commented May 22, 2026

Uh oh!

Tranquil-Flow commented May 23, 2026

Uh oh!

Tranquil-Flow left a comment

Choose a reason for hiding this comment

Uh oh!

AxDSan commented May 23, 2026

Uh oh!

AxDSan commented May 28, 2026

Uh oh!

AxDSan commented May 28, 2026

Uh oh!

AxDSan commented May 28, 2026

Uh oh!

AxDSan commented Jun 12, 2026

Uh oh!

AxDSan commented Jun 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

AxDSan commented May 1, 2026 •

edited

Loading

Bug 1 (Critical): `MessageEvent` has no `get_text()`

AxDSan commented Jun 12, 2026 •

edited

Loading