AgentSkillsCN

novita-docs

Novita AI 平台参考手册,涵盖 API、基础设施与集成相关任务。适用于用户咨询 Novita LLM API 的使用方法(兼容 OpenAI 的端点)、模型选型、Agent Sandbox 代码执行、GPU 实例或无服务器 GPU 部署、框架与工具集成、身份认证、计费方式、定价策略、配额/速率限制,以及故障排查等场景时使用。对于“如何将 Novita 与 X 结合使用?”和“为什么我的 Novita 请求会失败?”这类问题,优先推荐此技能。

SKILL.md
--- frontmatter
name: novita-docs
description: |
  Novita AI platform reference for API, infrastructure, and integration tasks. Use when users ask about Novita LLM API usage (OpenAI-compatible endpoint), model selection, Agent Sandbox code execution, GPU Instance or Serverless GPU deployment, framework/tool integrations, authentication, billing, pricing, quota/rate limits, or troubleshooting. Prefer this skill for "How do I use Novita with X?" and "Why is my Novita request failing?" questions.

Novita AI Platform Reference

Follow this workflow:

  1. Identify the user's product area (LLM, Sandbox, GPU, integration, troubleshooting).
  2. Open only the minimum reference file from the map below.
  3. For prices, model availability, and status, fetch live data before answering.
  4. Return copy-pasteable commands or code snippets whenever possible.
  5. If requested details are not covered in local references, provide the closest verified guidance, state the gap explicitly, and point to the official live docs.

Quick Reference

KeyValue
API Base URLhttps://api.novita.ai/openai
AuthenticationAuthorization: Bearer <API_KEY>
Get API Keyhttps://novita.ai/settings/key-management
Model Cataloghttps://novita.ai/models
Pricinghttps://novita.ai/pricing
Consolehttps://novita.ai/console
SupportDiscord: https://discord.gg/YyPRAzwp7P

Documentation Map

QuestionRead ThisLive Docs
Getting startedquick-start.md-
LLM API usagellm-guide.mdAPI Ref
Agent Sandboxsandbox-guide.mdDocs
GPU instancesgpu-guide.mdDocs
Tool integrationsintegrations.md-
Common issuescommon-issues.mdFAQ

Reference Selection Rules

  • Use references/quick-start.md for first-call setup and minimal examples.
  • Use references/llm-guide.md for chat completions, tools, vision, JSON mode, and batch jobs.
  • Use references/sandbox-guide.md for SDK/CLI, lifecycle, templates, and file operations.
  • Use references/gpu-guide.md for choosing between GPU Instance vs Serverless GPU and pricing tradeoffs.
  • Use references/integrations.md as the only integrations entrypoint; select sub-documents from that router.
  • Use references/common-issues.md first when debugging.
  • If the answer includes quotas, pricing, model counts, latency, or hardware availability, verify with live docs/API first and state the verification date.

Dynamic Data Policy

  • Treat model catalog size, pricing, rate limits, latency, and regional availability as dynamic.
  • Verify dynamic values with live docs/API before answering.
  • Report dynamic values with Verified on: YYYY-MM-DD in final responses.
  • If live verification is unavailable, explicitly say data may be stale and link the official source.

Dynamic Data (Always Fetch Live)

Get Latest Model List

bash
curl https://api.novita.ai/openai/v1/models \
  -H "Authorization: Bearer $NOVITA_API_KEY"

Current Pricing

Visit https://novita.ai/pricing for up-to-date pricing.

Changelog

Visit https://novita.ai/docs/changelog for latest updates.