dsh-rate-limiter

A proactive rate limiter plugin for DeepSeek Harness (dsh) that controls request rate per provider with a token bucket before model requests are issued, and queues over-limit requests with a delay instead of failing, to avoid upstream 429s. It is described as complementing dsh-llm-retry, which handles backoff after failure.

• Per-provider token bucket enforced before the request is sent (proactive prevention)
• Over-limit requests are queued with a delay instead of rejected, avoiding upstream 429s
• Unconfigured providers pass through untouched (zero intrusion)
• Queued waits honor the abort signal: stopping the user interrupts the wait immediately
• Hand-written reservation-based token bucket described as concurrency-safe, with zero third-party rate-limiting dependencies
• Mounts on agent/request and coexists with dsh-llm-retry (which mounts on agent/request-error)
• Documented as never modifying request content, changing routing, or swallowing errors — only controlling when a request is issued
• Configured per provider with rate (tokens/second) and burst (bucket capacity); enabled: false disables the plugin