Agent Skills
› NeverSight/learn-skills.dev
› vllm-studio-backend
vllm-studio-backend
GitHub指导处理vLLM Studio后端架构，涵盖控制器运行时、Pi-mono智能体循环、OpenAI兼容端点、LiteLLM网关及推理流程。用于修改路由、调试端点行为、更新智能体运行时及理解组件集成。
Trigger Scenarios
修改控制器路由或流式传输逻辑
调试OpenAI兼容端点行为
更新Pi-mono智能体运行时或工具执行
理解推理与LiteLLM的集成方式
Install
npx skills add NeverSight/learn-skills.dev --skill vllm-studio-backend -g -y
SKILL.md
Frontmatter
{
"name": "vllm-studio-backend",
"description": "Use when working on vLLM Studio backend architecture (controller runtime, Pi-mono agent loop, OpenAI-compatible endpoints, LiteLLM gateway, inference process, and debugging commands)."
}
vLLM Studio Backend Architecture
Overview
This skill explains how the backend is wired: controller runtime, OpenAI-compatible proxy, Pi-mono agent loop, LiteLLM gateway, and inference process management.
When To Use
- Modifying controller routes or run streaming.
- Debugging OpenAI-compatible endpoint behavior.
- Updating Pi-mono agent runtime or tool execution.
- Understanding how inference + LiteLLM fit together.
Quick Start
- Read
references/backend-architecture.mdfor the component map and data flow. - Read
references/openai-compat.mdfor/v1/modelsand/v1/chat/completionsbehavior. - Read
references/backend-commands.mdfor useful run/debug commands.
Core Guarantees
- Keep OpenAI-compatible endpoints stable (
/v1/models,/v1/chat/completions). /chatUI uses controller run stream (/chats/:id/turn) and Pi-mono runtime.- Tool execution happens server-side (MCP + AgentFS + plan tools).
References
references/backend-architecture.mdreferences/openai-compat.mdreferences/backend-commands.md
Version History
- e0220ca Current 2026-07-05 20:37
