Handle according to the following process:
- Identify the user's product area (model services, Agent Sandbox, GPU containers, integrations, troubleshooting).
- Use WebFetch or an equivalent HTTP reading tool to retrieve the minimum required reference files from the mapping table below.
- Treat pricing, model availability, and service status as dynamic data; obtain the latest information before answering.
- Whenever possible, return commands or code snippets that can be copied and pasted directly.
- If the local reference files do not cover the required content, provide the closest verified guidance, clearly state the gap, and point to the official live documentation.
LLM (Large Language Model) Quick Reference
| Key | Value |
|---|---|
| API Base URL | https://api.ppio.com/openai |
| Authentication method | Authorization: Bearer <API_KEY> |
| Get API Key | https://ppio.com/settings/key-management |
| Model catalog | https://ppio.com/ai-computing/llm-api |
| Pricing | https://ppio.com/pricing |
| Console | https://ppio.com/console |
Documentation mapping table
| Question | Read this file | Live documentation |
|---|---|---|
| Quick start | quick-start.md | - |
| LLM API usage | llm-guide.md | API reference |
| Agent Sandbox | sandbox-guide.md | Documentation |
| GPU instances | gpu-guide.md | Documentation |
| Tool integrations | integrations.md | Documentation |
| FAQ | faq-and-issues.md | FAQ |
Reference file selection rules
- For initial configuration and minimal examples, use
quick-start.md(see the mapping table above). - For chat completions, tool calling, vision, JSON mode, and batch tasks, use
llm-guide.md(see the mapping table above). - For SDK/CLI, lifecycle, templates, and file operations, use
sandbox-guide.md(see the mapping table above). - For choosing between GPU instances and serverless GPU and for pricing trade-offs, use
gpu-guide.md(see the mapping table above). - For integration issues, use
integrations.md(see the mapping table above) as the unified entry point, then select sub-documents from it. - When debugging, prioritize using
faq-and-issues.md(see the mapping table above). - If the raw files cannot be retrieved from GitHub, fall back to the local
references/*.mdfiles in this skill directory. - If the answer involves quotas, pricing, model counts, latency, or hardware availability, verify through live documentation/API first, and note the verification date in the response.
Dynamic data policy
- Treat model catalog size, pricing, rate limiting, latency, and regional availability as dynamic data.
- In the final response, mark the verification time of dynamic data in the
Verified on: YYYY-MM-DDformat. - If real-time verification is not possible, clearly state that the data may be outdated and include links to official sources.
Dynamic data (always retrieve the latest values)
Get the latest model list
curl https://api.ppio.com/openai/v1/models \
-H "Authorization: Bearer $PPIO_API_KEY"
Current pricing
Visit https://ppio.com/pricing to view the latest pricing.
Changelog
- Model API Service release notes: https://ppio.com/docs/announcement/changelog-llm
- GPU Container Service release notes: https://ppio.com/docs/announcement/changelog-gpu
- Agent Sandbox release notes: https://ppio.com/docs/announcement/changelog-agent-sx