OpenTelemetry tracing for DeepSeek Harness (dsh): turns each agent turn into a GenAI span tree — steps, LLM calls with TTFT, tool executions, token usage — exported over standard OTLP to Jaeger, Grafana Tempo, SigNoz, Langfuse, or any compatible backend.
Zero-touch runtime telemetry for DeepSeek Harness: every session writes its own report card (task/model/tools/failures/duration) into a human-readable journal. 纯观察者运行事实遥测。
Per-round reasoning_effort optimizer for DeepSeek Harness (dsh): auto-downgrades tool-call reasoning for simple tool chains, lifting back for heavy work. Cuts thinking time between tool calls.
Developer tool: live token usage & cost monitoring for DeepSeek Harness - consumed tokens for the current session and across all sessions, read from token-meter projections. No model calls.
Cross-session persistent cost ledger for DeepSeek Harness: logs every LLM token usage to SQLite and exposes record/query/budget tools. Built-in DeepSeek pricing, overridable via config.
dsh-polling — 轮询任务/定时任务 plugin for DeepSeek Harness: cron scheduled tasks as real sessions, natural-language creation, model tools (polling_*) & web UI. Install via dsh plugin add
Delayed wake-up for DeepSeek Harness: schedule a reminder and the agent wakes in the SAME conversation to reply — plus an interruptible wait (timed gate) tool. Host-level timer + official agent.followup() wake channel.