fix(gateway): show iteration progress and idle time in /status

[LifeJiggy]

When an agent is running, /status now shows:

• Iteration progress (e.g. 'iter 3/50')
• Idle time since last activity (e.g. 'idle 2s')
• Last activity description (e.g. 'calling web_search')

For pending agents (starting up), shows 'starting up' instead.

Enhancements:

• Imported _AGENT_PENDING_SENTINEL for clean sentinel detection
• Graceful fallback when agent lacks get_activity_summary
• Error handling prevents status command from failing on bad agent state

Tests:

• test_status_shows_iteration_progress_when_agent_running
• test_status_shows_startup_when_agent_pending
• test_status_no_detail_when_no_activity_summary

What does this PR do?

Enhances /status command output in the gateway to show real-time agent progress: iteration count (iter 3/50), idle time (idle 2s), and last activity description. Also handles the _AGENT_PENDING_SENTINEL state (agent starting up) with a distinct label instead of showing "unknown".

When the agent is running, /status now shows a running_detail line with iteration progress, idle time, and last tool/activity description. When the agent is pending (starting up), it shows "starting up" instead of "unknown". Falls back gracefully when get_activity_summary is unavailable.

Related Issue
Enhances observability for long-running agent sessions

Type of Change

• ✨ New feature (non-breaking change that adds functionality)

Changes Made

• gateway/slash_commands.py: Added running_detail computation in _handle_status_command with iteration progress, idle time, and last activity description; _AGENT_PENDING_SENTINEL gets special "starting up" label
• tests/gateway/test_status_command.py: 3 new tests covering iteration progress display, pending sentinel handling, and no-summary fallback

How to Test

1. python -m pytest tests/gateway/test_status_command.py -q — 17/17 pass
2. In a running gateway session, send /status and verify iteration/idle/detail line appears
3. Verify "starting up" label when agent is in pending state

Checklist

• I've read the Contributing Guide
• My commit messages follow Conventional Commits
• I searched for existing PRs
• My PR contains only changes related to this fix
• I've run pytest and all tests pass
• I've added tests for my changes
• I've tested on Windows 11

[tonydwb]

Code Review Summary

Verdict: Approved

Looks Good

• gateway/slash_commands.py:9-23 — When agent is running, calls get_activity_summary() on the running agent and formats iteration count (api_call_count/max_iterations), idle time (seconds_since_activity), and last activity description into the status line. Uses local import to avoid circular dependency on _AGENT_PENDING_SENTINEL.
• Handles three states gracefully: pending sentinel (starting up), normal agent with get_activity_summary, and agent without the method (gracefully omitted).
• tests/gateway/test_status_command.py — Three new tests covering: iteration progress display, startup sentinel, and graceful omission when method is absent.

No Issues Found

• Exception handling around get_activity_summary() is appropriate — silently skips if the method throws.
• Idle time rounding to integer seconds is correct for human-readable output.

---

Reviewed by Hermes Agent

[teknium1]

Thanks for adding useful runtime observability. The underlying data is available on current main (run_agent.py:3297-3313), but the PR needs adaptation to the current status handler.

Problems

• Current main deliberately treats _AGENT_PENDING_SENTINEL as not running (gateway/slash_commands.py:505-510). Because the PR puts its starting up handling under if is_running, that branch would not execute after resolving the current conflict.
• The PR's new suffixes are English-only, while current /status uses localized templates (gateway/slash_commands.py:627-645) and the status schema is defined in locales/en.yaml:285-305 and the other locale files.

Suggested changes

• Separate pending-slot detection from live-agent detection, then cover the sentinel case against the current handler.
• Add translation keys/placeholders for the progress detail rather than concatenating English text.

Automated hermes-sweeper review.

[LifeJiggy]

Addressed both points from your review:

1. Separated pending-slot detection from live-agent detection — added is_pending = agent is _AGENT_PENDING_SENTINEL before the is_running check at slash_commands.py:510. The original code put sentinel handling under if is_running, which excluded it since is_running explicitly filters out the sentinel. Now the pending state is detected and handled independently.

2. Added translation keys for progress detail — added state_pending and state_running_detail keys to all 16 locale files (en, de, es, fr, ga, hu, it, ja, ko, pt, ru, tr, uk, zh, zh-hant, af). Progress detail (iter count, idle time, last activity) uses localized {detail} placeholder instead of English-only string concatenation.

Changes:

• gateway/slash_commands.py: separate is_pending/is_running detection, build agent_state with localized templates, fallback to state_yes on error
• locales/*.yaml: add state_pending and state_running_detail keys
• tests/gateway/test_status_command.py: 4 new tests (iteration progress, pending sentinel, no activity summary, sentinel not treated as running)

21/21 status command tests pass.