Read-only performance diagnostics for DeepSeek Harness: session load/restore timing, spill-hit counts, compaction count and trigger, context-injection volume (AGENTS.md/skills/tool-schema token share), and LLM cache hit rate — surfaced via /fast, persisted as reconstructable session events with async sampling off the model path.
Cross-session usage & cost observatory for the DeepSeek Harness web profile — trend/heatmap dashboards, per-model peak-hour pricing (CNY/USD), official DeepSeek balance with spend reconciliation.