feat(status): show model, context usage, and cumulative label in /status

[WolframRavenwolf]

Summary

Enhances the /status slash command in the gateway to display:

• Model: name and provider (e.g., "claude-opus-4-20250514 (anthropic)")
• Context: approximate token usage vs limit with percentage (e.g., "~45,000 / 200,000 (23%)")
• Tokens: now labeled "(cumulative)" for clarity

Motivation

Users had no way to see which model was active or how much context was consumed without checking config files or logs. This information is critical for understanding agent behavior and knowing when context compression might kick in.

Changes

gateway/run.py: Enhanced _handle_status_command with:

• Live agent path: reads model, provider, context usage from the running agent instance and its context_compressor
• Idle fallback: resolves model from config via _resolve_gateway_model(), context from session entry + DEFAULT_CONTEXT_LENGTHS fuzzy lookup
• Correct sentinel check (agent is not _AGENT_PENDING_SENTINEL) — pending agents now correctly report as "not running", consistent with _handle_stop_command
• Zero-context handling: context line shows even when context_used=0 (e.g., after /new) — only context_total is checked, avoiding the Python falsy-zero pitfall

Example Output

📊 Hermes Gateway Status

Session ID: a1b2c3d4e5f6...
Created: 2026-03-31 14:30
Last Activity: 2026-03-31 15:45
Model: claude-opus-4-20250514 (anthropic)
Context: ~45,231 / 200,000 (23%)
Tokens: 125,432 (cumulative)
Agent Running: Yes ⚡

Connected Platforms: telegram, discord, whatsapp

Design Decisions

• No new config options — this is purely additive information display
• Graceful degradation — if model isn't in DEFAULT_CONTEXT_LENGTHS, context line is simply omitted rather than showing incorrect data
• Consistent patterns — reuses existing _load_gateway_config() and _resolve_gateway_model() for the idle path, same as /model and agent spawn
• Minimal diff — single file, ~57 lines added, focused change

Test Plan

• Model line appears when agent is running (reads from agent.model and agent.provider)
• Context line shows "~0 / 200,000 (0%)" after /new (zero-context edge case)
• Graceful degradation when model not in DEFAULT_CONTEXT_LENGTHS
• Idle fallback resolves model from config and context from session entry
• Pending agents (sentinel) correctly report "Agent Running: No"

[malaiwah]

Nice work on the idle context fallback — PR #8355 incorporates the same idea (using get_model_context_length() when no live agent exists) as part of a broader /status overhaul that also adds token breakdown, cost, cache stats, compression tracking, and CLI support. Credit to @WolframRavenwolf for the approach.

[alt-glitch]

Related to #16079 (also adds model/provider info to /status) — these likely overlap significantly.

[alt-glitch]

Related to #16079