dsh-agent-rate-limit

@zhourenke/dsh-agent-rate-limit

A DSH agent loop rate limiter that prevents TPM (Tokens Per Minute) and RPM (Requests Per Minute) limit violations by intercepting the LLM streaming pipeline and adding adaptive delays between requests. Also retries HTTP 429 responses with escalating backoff.

Installation

Copy the package to your DSH profile's node_modules:

# From the repo root
Copy-Item -Recurse -Force .\dsh-agent-rate-limit $env:USERPROFILE\.dsh\profiles\web\node_modules\@zhourenke\dsh-agent-rate-limit

Or use a directory junction for development:

New-Item -ItemType Junction -Path "$env:USERPROFILE\.dsh\profiles\web\node_modules\@zhourenke\dsh-agent-rate-limit" -Target "C:\path\to\dsh-agent-rate-limit"

Usage

Add to your agent preset:

- id: agent-rate-limit
  name: '@zhourenke/dsh-agent-rate-limit'

Configuration

Key	Default	Description
windowMs	60000	Sliding window size in milliseconds (60s).
tpmLimit	1200000	TPM (Tokens Per Minute) limit. Default matches Alibaba Cloud Bailian deepseek-v4-flash.
rpmLimit	15000	RPM (Requests Per Minute) limit.
safetyFactor	0.8	Safety factor (0.8 = use 80% of the limit, leaving 20% buffer).
maxBackoffMs	30000	Maximum backoff delay in milliseconds (30s).
retryOn429	true	When true, HTTP 429 responses are silently retried with escalating backoff.
maxRetries	5	Maximum consecutive 429 retries per burst before giving up.
verbose	false	When true, log per-request details (delay, token recording, retry attempts).

Check status

Type /agent-rate-limit in the chat input to see current plugin status and configuration.

How it works

Token tracking

1. Token recording: After each successful model call, the plugin records the exact token count from the API's usage chunk — far more accurate than heuristic estimation.
2. Sliding window: A 60-second FIFO queue tracks recent token consumption. Before each request, the plugin checks if the current window is approaching the TPM or RPM limit and delays accordingly.
3. Input estimation: For the pre-request delay calculation, the plugin uses the average of the last 3 actual input token counts from the API. Falls back to heuristic estimation (CJK: ~1.5 chars/token, others: ~3.5 chars/token) only for the very first request.

Error recovery

When a rate-limit error is detected (HTTP 429), the plugin:

1. Increments the retry count for the current burst
2. Returns { kind: 'retry' } to retry transparently
3. Applies escalating backoff: 2s → 4s → 8s → 16s → 30s (capped at maxBackoffMs)
4. Gives up after maxRetries consecutive failures and surfaces the error
5. Resets the retry counter on success

Rate limit detection

The plugin detects HTTP 429 by checking the error's statusCode, code, and message for patterns like rate limit, too many requests, tpm, rpm, quota, throttl, and 429.

Credits

Built for DeepSeek Harness.