fix(cli): dispatch /agents inline while agent is running (#32477) - #32541
Closed
briandevans wants to merge 2 commits into
Closed
fix(cli): dispatch /agents inline while agent is running (#32477)#32541briandevans wants to merge 2 commits into
briandevans wants to merge 2 commits into
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
Note
Copilot was unable to run its full agentic suite in this review.
Adds inline (busy-path) dispatch for the classic CLI /agents command (and /tasks alias) so it runs immediately during an active agent loop instead of being queued behind process_loop and effectively doing nothing until the run completes.
Changes:
- Add
_should_handle_agents_command_inline()to detect when/agentsor/tasksshould bypass the queue while the agent is running - Update
handle_enter()to print and dispatch/agentsinline when busy - Add regression tests validating detector behavior and that
/agents//tasksdo not enqueue whenprocess_command()is invoked
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 6 comments.
| File | Description |
|---|---|
| tests/cli/test_cli_agents_busy_path.py | Adds regression tests for inline detection/dispatch for /agents and /tasks while busy |
| cli.py | Implements the inline detector and triggers inline dispatch from the UI thread in handle_enter() |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| @@ -7792,6 +7792,30 @@ def _should_handle_steer_command_inline(self, text: str, has_images: bool = Fals | |||
| except Exception: | |||
| return False | |||
|
|
|||
| def _should_handle_agents_command_inline(self, text: str, has_images: bool = False) -> bool: | |||
Comment on lines
+7807
to
+7817
| if not text or has_images or not _looks_like_slash_command(text): | ||
| return False | ||
| if not getattr(self, "_agent_running", False): | ||
| return False | ||
| try: | ||
| from hermes_cli.commands import resolve_command | ||
| base = text.split(None, 1)[0].lower().lstrip('/') | ||
| cmd = resolve_command(base) | ||
| return bool(cmd and cmd.name == "agents") | ||
| except Exception: | ||
| return False |
Comment on lines
+7811
to
+7814
| try: | ||
| from hermes_cli.commands import resolve_command | ||
| base = text.split(None, 1)[0].lower().lstrip('/') | ||
| cmd = resolve_command(base) |
| # makes it useless: the user sees "nothing happens" until the | ||
| # whole delegation chain finishes. See #32477. | ||
| if self._should_handle_agents_command_inline(text, has_images=has_images): | ||
| _cprint(f"\n⚙️ {text}") |
Comment on lines
+113
to
+138
| """When the detector fires, /agents dispatch through process_command must | ||
| reach _handle_agents_command directly rather than being queued.""" | ||
|
|
||
| def test_process_command_routes_to_agents_handler(self): | ||
| """With _agent_running=True, /agents calls _handle_agents_command and | ||
| does NOT enqueue onto _pending_input.""" | ||
| cli = _make_cli() | ||
| cli._agent_running = True | ||
| cli._pending_input = MagicMock() | ||
| cli._handle_agents_command = MagicMock() | ||
|
|
||
| cli.process_command("/agents") | ||
|
|
||
| cli._handle_agents_command.assert_called_once() | ||
| cli._pending_input.put.assert_not_called() | ||
|
|
||
| def test_tasks_alias_routes_to_agents_handler(self): | ||
| """/tasks resolves to canonical 'agents' and dispatches the same.""" | ||
| cli = _make_cli() | ||
| cli._agent_running = True | ||
| cli._pending_input = MagicMock() | ||
| cli._handle_agents_command = MagicMock() | ||
|
|
||
| cli.process_command("/tasks") | ||
|
|
||
| cli._handle_agents_command.assert_called_once() |
Comment on lines
+113
to
+138
| """When the detector fires, /agents dispatch through process_command must | ||
| reach _handle_agents_command directly rather than being queued.""" | ||
|
|
||
| def test_process_command_routes_to_agents_handler(self): | ||
| """With _agent_running=True, /agents calls _handle_agents_command and | ||
| does NOT enqueue onto _pending_input.""" | ||
| cli = _make_cli() | ||
| cli._agent_running = True | ||
| cli._pending_input = MagicMock() | ||
| cli._handle_agents_command = MagicMock() | ||
|
|
||
| cli.process_command("/agents") | ||
|
|
||
| cli._handle_agents_command.assert_called_once() | ||
| cli._pending_input.put.assert_not_called() | ||
|
|
||
| def test_tasks_alias_routes_to_agents_handler(self): | ||
| """/tasks resolves to canonical 'agents' and dispatches the same.""" | ||
| cli = _make_cli() | ||
| cli._agent_running = True | ||
| cli._pending_input = MagicMock() | ||
| cli._handle_agents_command = MagicMock() | ||
|
|
||
| cli.process_command("/tasks") | ||
|
|
||
| cli._handle_agents_command.assert_called_once() |
alt-glitch
added
P2
Contributor
Author
|
@copilot Addressed in commit 3b5eb124d:
The CI failure on |
…h#32477) Typing /agents (or its /tasks alias) during an active agent turn does nothing in classic CLI: the slash command goes through _pending_input, which process_loop only drains after self.chat() returns. By the time the queued command is pulled, the user has been staring at a quiet screen for the entire delegation chain — defeating the introspection command's only purpose. Mirror the inline-dispatch pattern PR NousResearch#25011 introduced for /steer: when the detector sees /agents while the agent is running, call process_command directly on the UI thread. _handle_agents_command is read-only (reads process_registry + a couple of CLI attrs) and _cprint already routes thread-unsafe output through prompt_toolkit's run_in_terminal, so UI-thread dispatch is safe. Fixes NousResearch#32477
Address Copilot inline review on NousResearch#32541: * Reject `/agents foo` / `/tasks something` on the busy path — `/agents` takes no args, so trailing tokens must fall through to the normal dispatch where `_handle_agents_command` can surface a usage error rather than silently dropping the argument. * Drop the `⚙️ {text}` echo from the busy-path block. Mirrors the existing `/steer` block which dispatches silently and lets `_handle_agents_command`'s own output land via `_cprint` / `run_in_terminal`. Avoids the screen-reader / terminal-font concern Copilot flagged. Adds a `test_ignores_agents_with_trailing_args` case covering the new detector constraint.
briandevans
force-pushed
the
fix/cli-agents-inline-busy-path-32477
branch
from
May 28, 2026 17:09
3b5eb12 to
b9d2f4a
Compare
Contributor
Author
|
Closing to focus the queue on security/file-safety work where civilian merges are landing. Happy to reopen if maintainers want this picked up. |
1 task
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Typing
/agents(or its/tasksalias) during an active agent turn in the classic CLI does nothing — the user reports "the agents appear to be delegating properly, but attempting to monitor with /tasks or /agents does nothing". The slash command flows through_pending_input, whichprocess_looponly drains afterself.chat()returns; until then the queued command sits silently behind the running delegation chain. From the user's perspective the keystroke is lost during exactly the window the command exists for.Mirror the inline-dispatch pattern PR #25011 introduced for
/steer: detect/agentson the UI thread insidehandle_enterand callprocess_commanddirectly._handle_agents_commandis read-only (touchesprocess_registryplus a couple of CLI attrs) and_cprintalready routes thread-unsafe output throughprompt_toolkit'srun_in_terminal, so UI-thread dispatch is safe.Related Issue
Fixes #32477
Type of Change
Changes Made
cli.py— add_should_handle_agents_command_inline(parallel to_should_handle_steer_command_inline) and dispatch inline fromhandle_enterwhen the detector fires.tests/cli/test_cli_agents_busy_path.py— new regression suite mirroringtest_cli_steer_busy_path.py: detector returns True for/agentsand/taskswhile running, False when idle / non-slash / images attached / other slash commands;process_commandroutes to_handle_agents_commandwithout enqueueing on_pending_input.How to Test
hermes chatwith a delegating profile./agents(or/tasks). The active-agents + running-processes summary should print above the prompt immediately instead of disappearing into the queue.uv run --with pytest --with pytest-xdist --with pytest-asyncio python3 -m pytest tests/cli/test_cli_agents_busy_path.py tests/cli/test_cli_steer_busy_path.py -v— 15 passes.Checklist
Code
fix(scope):,feat(scope):, etc.)Documentation & Housekeeping
Screenshots / Logs