agentsclimarketplace

Solon ai

Skill chengliang4810/jimuqu-skills/skills/solon-ai

Install
npx -y skills add chengliang4810/jimuqu-skills --skill solon-ai

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

辅助开发、迁移、调试和解释 Solon AI Java AI 应用。只要用户提到 Solon AI、solon-ai、ChatModel、EmbeddingModel、RAG、DocumentLoader、Repository、向量库、ToolMapping、Skill、MCP、McpServerEndpoint、Agent、ReActAgent、TeamAgent、AI Flow、provider 方言、Ollama/OpenAI/DashScope/Gemini/Anthropic 接入,或需要根据 Solon AI 官方开源仓库确认 API 与机制,都必须使用此技能。

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.6 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it

Solon AI Dev Skill

你在辅助 Solon AI 项目开发。Solon AI 的 API 变化快,公开资料有限,回答和实现必须优先以目标项目依赖版本和官方源码为准。

官方仓库:https://gitee.com/opensolon/solon-ai

首要原则

  1. 先判定任务类型:模型接入、配置注入、工具调用、RAG、向量库、MCP、Agent、Flow、问题排查、迁移改造。
  2. 优先查证源码:不确定的类、注解、配置键、provider、模块依赖,必须读源码或测试示例。
  3. 不要套用 Spring AI / LangChain 习惯:Solon AI 有自己的 ChatModelEmbeddingModelToolMappingRepositoryMcpServerEndpoint、Agent API。
  4. 输出要可落地:给 Java 代码、app.yml 配置、依赖模块选择、源码依据和验证方式。
  5. 安全边界:涉及工具调用、文件/命令/浏览器 skill、MCP 暴露时,提醒权限、输入校验和可见副作用。

资料入口

按任务读取对应 reference,必要时再读源码:

  • 仓库结构和模块选择:references/overview-modules.md
  • 配置与注入:references/configuration-and-injection.md
  • ChatModel 与 provider 方言:references/chat-model-and-dialects.md
  • 工具调用与 Skills:references/tool-calling-skills.md
  • RAG 主链路:references/rag-load-split-retrieve.md
  • 向量库模块选择:references/vector-repositories.md
  • MCP 客户端/服务端:references/mcp-client-server.md
  • Agent:references/agent-react-team.md
  • AI Flow:references/flow-yaml.md
  • 示例与测试索引:references/tests-and-examples-map.md

如果任务很具体,直接 grep 源码优先于泛读 reference。

工作流

开发功能

  1. 查当前项目已有 Solon AI 用法和依赖。
  2. 读取对应 reference,确认模块和 API 名称。
  3. 必要时读取 Solon AI 测试示例,优先复用真实写法。
  4. 给最小实现,不引入未请求的编排层或多 provider 抽象。
  5. 验证:至少编译;如涉及模型调用,说明需要真实 apiKey/本地模型服务。

问题排查

  1. 收集异常、配置、依赖模块、provider、apiUrl、model。
  2. 对照源码确认失败点:方言选择、配置绑定、工具注册、MCP endpoint、向量库连接或 loader 依赖。
  3. 给最小修复和验证命令。

源码机制解释

  1. 定位源码定义和插件声明。
  2. 读取调用链和测试示例,不只读 README。
  3. 输出:结论 → 关键源码路径 → 机制流程 → 坑点。

常用查证命令

# 核心类
rg "class ChatModel|class EmbeddingModel|@interface ToolMapping" solon-ai-core/src/main/java

# 插件声明
find . -path '*/META-INF/solon/*.properties'

# 测试示例
find . -path '*/src/test/java/*' -type f | grep 'features/ai\|demo/ai'

输出规范

  • 全程中文;Java 类名、配置键和代码标识保持原样。
  • 涉及源码机制时引用 path:line
  • 不确定时说“源码中未确认”,不要补 Spring AI 或 LangChain 的 API。
  • 代码示例默认使用 Solon 注解,不使用 Spring 注解,除非用户明确在 SpringBoot 嵌入场景。

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.