1. 账户邮箱 (Account Email)：chengzi9988@gmail.com。
2. API Key 后四位 (Last 4 digits)：JXrD3
3. 当前限额 (Current Limit)：40 RPM。
4. 申请额度 (Requested Limit)： 200 RPM。
5. 申请理由 (Use Case)：Hello NVIDIA Support Team,
I am a developer working on an open-source AI agent framework called Hermes Agent (https://github.com/NousResearch/hermes-agent). My current project involves building a multi-step reasoning agent that leverages NVIDIA NIM APIs for both LLM inference and embedding generation.
Unlike simple single-turn Q&A applications, my agent performs complex, multi-step workflows – it iteratively reasons, calls external tools (such as web search, code execution, and database queries), and synthesizes results over several cycles. In practice, a single user query can easily trigger 20 to 50 individual API calls to the NIM service (including both chat completions and embedding requests).
The current 40 RPM rate limit frequently results in HTTP 429 errors during normal testing, which interrupts the workflow and significantly hinders my development progress. I am not running a commercial production service; this is strictly for personal development, testing, and open-source contribution purposes.
To facilitate smoother testing and allow me to validate the agent’s performance under more realistic conditions, I kindly request an increase to 200 RPM.
Thank you for your consideration. I appreciate the work you do in providing these powerful services to the developer community. 您好，NVIDIA 支持团队：
我是一名开发者，目前正在基于开源 AI Agent 框架 Hermes Agent（https://github.com/NousResearch/hermes-agent）进行项目开发。我的工作内容是构建一个多步推理智能体，该智能体同时依赖 NVIDIA NIM API 进行大语言模型推理和向量嵌入生成。
与简单的单轮问答应用不同，我的智能体执行的是复杂的多步工作流——它会反复进行推理、调用外部工具（如网络搜索、代码执行和数据库查询），并在多个循环中整合结果。在实际运行中，单次用户提问就可能触发 20 到 50 次对 NIM 服务的 API 调用（包括对话补全和嵌入请求）。
当前的 40 RPM 速率限制在正常测试中频繁导致 HTTP 429 错误，中断工作流，严重阻碍了我的开发进度。我并非在运行商业生产服务，我的使用场景严格限于个人开发、测试以及开源贡献。
为了更顺畅地进行测试，并让我能够在更接近真实场景的条件下验证智能体的性能，我诚恳地申请将限额提升至 200 RPM。
感谢您的考虑，也感谢你们为开发者社区提供如此强大的服务。