Expose context/compression state in gateway-visible status command

[oqpsk]

Problem

OpenClaw's /status exposed practical session-state visibility, but Hermes on gateway platforms (especially Discord) does not currently expose the pieces that matter most for long-running chats.

For Discord-first users, CLI-only affordances are not a real workaround. /usage helps with token usage, but it does not answer the operational question:

• what is the current context window?
• how full is it?
• has context compression already happened?
• how many times?
• when was the last compression?

Right now, users can infer that compression exists, but cannot reliably observe whether it happened in the current session.

Requested improvement

Expose session context/compression state in a gateway-visible command.

Two reasonable options:

1. Add a new /context command
2. Expand existing /status (preferred if you want fewer commands)

Minimum useful fields

• current model
• current provider
• context window
• estimated context usage %
• compression happened: yes/no
• compression count
• last compression time

Why this matters

This is not just a nice-to-have metric display. It affects trust and usability in Discord/Telegram/Slack sessions where users do not live in the CLI and want to know whether the agent is still operating on full context or already running on compressed history.

Notes

• /usage is adjacent, but insufficient
• CLI banner/statusbar already surface some context-window information, so this is mainly a gateway observability gap
• Discord-first workflows should not require switching to CLI just to answer “did this session already compress?”

[Dille68]

Cron update:

Root cause:

• gateway/run.py:_handle_status_command() only returned basic session metadata (ID/title/timestamps/tokens/running/platforms).
• It did not expose model/provider/context-window usage or live compression state, so gateway users could not tell how full the session was or when reset/compact was needed.

Files changed:

• gateway/run.py
• tests/gateway/test_status_command.py

Diff summary:

• expanded /status to include model, provider, context window, context usage %, compression happened/count/last-compression, and reset guidance
• uses live context_compressor data when an agent is running
• falls back to configured/runtime model + persisted last_prompt_tokens when idle
• added regression tests for idle high-usage status and live compression-state reporting

Mini-test:

• source venv/bin/activate && python -m pytest tests/gateway/test_status_command.py -q
• Result: 6 passed in 0.82s

Evidence:

• paths: gateway/run.py, tests/gateway/test_status_command.py
• key hunks: /status now renders Model, Provider, Context Window, Context Usage, Compression Happened, Compression Count, Last Compression, Reset Guidance

Next:

• if this direction looks right, follow up by persisting compression metadata across completed runs so idle sessions can also show non-unknown compression history.

[malaiwah]

PR #8355 addresses this — it expands /status to show model, provider, context window usage (tokens/limit/%), compression count, cache hit rate, cost, and queue depth. Both gateway and CLI are covered. When no live agent is running, context is estimated from get_model_context_length() + last_prompt_tokens.

[alt-glitch]

Related to open PR #8355 (feat(status): rich /status snapshots across CLI and gateway) which claims to close this issue, and #10617 (add /context slash command).