Requesting RPM Increase (40 to 200 RPM) for Multi-Agent Workflows and RAG Developmen (original) (raw)

Hello NVIDIA Support Team,

I would like to request a rate limit increase for my NVIDIA NIM API account.

Account Details

• Email: montanarco.17104@gmail.com
• Current limit: 40 RPM
• Requested limit: 100–200 RPM (whatever is the next available tier for individual developers)

Use Case
I am using the NVIDIA NIM API for personal development, security auditing, and non-production testing of advanced multi-agent AI workflows. My setup focuses on local automation, system architecture, and specialized data engineering tasks.

A typical workflow involves:

• Multi-agent orchestration with parallel subagent calls to coordinate, route, and execute complex programming tasks.
• Application security checks, vulnerability scanning assistance, and secure code review using advanced code models.
• Data analytics pipelines, system design generation, and detailed infrastructure implementation plans.
• Building and evaluating local RAG (Retrieval-Augmented Generation) architectures that require frequent embedding and generation calls.

Models evaluated in this workflow:

• Qwen (including Qwen 2.5 Coder variants)
• Gemma 4
• Kimi 2.6 / Moonshot models and other high-context reasoning LLMs.

Why 40 RPM is Insufficient
Even with client-side throttling and exponential backoff, a single agentic execution loop (which plans, tests, reads files, and self-corrects) triggers dozens of concurrent API calls within a few seconds. The default 40 RPM global limit causes immediate 429 “Too Many Requests” errors, completely breaking the autonomous agent loops mid-execution.

Commitment

• Personal development and R&D only — no production or public-facing production services.
• Low concurrency usage, will continue respecting Retry-After headers.
• Full compliance with NVIDIA’s developer fair-use policy.

Thank you for considering my request.

Best regards,
Miguel