Agent skill · software engineering · jeremylongshore
ideogram-rate-limits
Control Ideogram concurrency with a bounded queue, deadlines, backoff, and fair tenant admission. Use when preventing 429 responses or sizing generation workers. Trigger with "tune Ideogram concurrency", "fix Ideogram throttling", or "design an Ideogram request queue".
Why this skill is useful
Adds executable scripts for handling Ideogram's API rate limits with backoff and queuing patterns that aren't in public docs.
What it needs
Requires p-queue installed locally. Requires ideogram account access. About 2k tokens when loaded. Last updated 2026-08-07. 2,604 stars on the source repository.
What this skill does
Ideogram Concurrency and Queue Control Overview Keep paid image work inside a measurable concurrency envelope. Combine admission control, tenant fairness, operation deadlines, and endpoint-aware retry so a traffic burst does not become duplicate spend, an unbounded queue, or persistent 429 pressure. Prerequisites Request arrival rate, latency objective, output count, endpoint mix, and cost ceiling. Queue ownership, per-tenant policy, cancellation behavior, and overload response. Current account capacity and an approved contact path for increased scale. Current Contract Ideogram's API overview documents a default limit of 10 in-flight requests and directs larger-capacity needs to partnership@ideogram.ai. This is a concurrency boundary, not permission to launch ten workers per process. Account-wide observed behavior remains authoritative. Authentication Workers inject IDEOGRAMAPIKEY server-side and send Api-Key only to https://api.ideogram.ai. Queue records and metrics must exclude keys, prompts, images, and response URLs. Instructions 1. Measure arrival rate, service time by endpoint, current in-flight count, queue age, timeout rate, and 429 responses. 2. Place a shared account-level semaphore below the verified limit; start conservatively and reserve capacity for recovery probes. 3. Add bounded per-tenant admission, queue length, age, output count, and total operation deadlines. 4. Prefer asynchronous endpoints for work that cannot fit an interactive deadline and persist each generationid. 5. Retry only classified transient responses using server guidance when present, exponential backoff with jitter, and a strict attempt ceiling. 6. Prevent ambiguous duplicate generation by reconciling known async identifiers before resubmission. 7. Raise capacity only through the vendor path and a reviewed load plan; canary the new limit. Tool Discipline Use Read, Glob, and Grep for queue, worker, metric, and fixture inspection. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills jeremylongshore/ideogram-rate-limits--246075