AI 写文章系列——Eino 介绍文档及 DeepSeek 示例

chain, _ := NewChain[map[string]any, *Message]().
           AppendChatTemplate(prompt).
           AppendChatModel(model).
           Compile(ctx)
chain.Invoke(ctx, map[string]any{"query": "what's your name?"})

graph := NewGraph[map[string]any, *schema.Message]()
_ = graph.AddChatTemplateNode("node_template", chatTpl)
_ = graph.AddChatModelNode("node_model", chatModel)
_ = graph.AddToolsNode("node_tools", toolsNode)
_ = graph.AddLambdaNode("node_converter", takeOne)
_ = graph.AddEdge(START, "node_template")
...

// 创建DeepSeek模型
deepSeekModel, err := deepseek.NewChatModel(ctx, &deepseek.ChatModelConfig{
    Model:   "deepseek-reasoner", // 或其他DeepSeek模型
    APIKey:  os.Getenv("DEEPSEEK_API_KEY"),
    BaseURL: os.Getenv("DEEPSEEK_BASE_URL"), // 可选
})

// 生成回复
message, err := deepSeekModel.Generate(ctx, []*schema.Message{
    schema.SystemMessage("你是一个有帮助的助手。"),
    schema.UserMessage("什么是Eino框架？")
})

// 获取推理内容（仅适用于支持推理的模型，如deepseek-reasoner）
if reasoning, ok := deepseek.GetReasoningContent(message); ok {
    fmt.Println("推理过程:", reasoning)
}

package main

import (
    "context"
    "fmt"
    "os"

    "github.com/cloudwego/eino/components/document"
    "github.com/cloudwego/eino/compose"
    "github.com/cloudwego/eino/schema"
    "github.com/cloudwego/eino-ext/components/document/loader/markdown"
    "github.com/cloudwego/eino-ext/components/document/transformer/splitter/recursive"
    "github.com/cloudwego/eino-ext/components/embedding/openai"
    "github.com/cloudwego/eino-ext/components/indexer/redis"
    "github.com/cloudwego/eino-ext/components/model/deepseek"
    "github.com/cloudwego/eino-ext/components/retriever/redis"
)

func main() {
    ctx := context.Background()
    
    // 1. 设置文档加载器
    loader, _ := markdown.NewLoader(ctx, &markdown.LoaderConfig{
        Source: document.Source{
            Path: "./知识库文档路径",
        },
    })
    
    // 2. 设置文档分割器
    splitter, _ := recursive.NewSplitter(ctx, &recursive.Config{
        ChunkSize:    1000,
        ChunkOverlap: 200,
    })
    
    // 3. 设置向量化模型
    embedder, _ := openai.NewEmbedding(ctx, &openai.EmbeddingConfig{
        APIKey: os.Getenv("OPENAI_API_KEY"),
        Model:  "text-embedding-3-large",
    })
    
    // 4. 设置索引器
    indexer, _ := redisindexer.NewIndexer(ctx, &redisindexer.IndexerConfig{
        Address:  "localhost:6379",
        Password: "",
        DBIndex:  0,
        IndexName: "eino-docs",
        Dimension: 3072, // 根据嵌入模型维度调整
    })
    
    // 5. 设置检索器
    retriever, _ := redisretriever.NewRetriever(ctx, &redisretriever.RetrieverConfig{
        Address:   "localhost:6379",
        Password:  "",
        DBIndex:   0,
        IndexName: "eino-docs",
        Embedding: embedder,
        TopK:      5,
    })
    
    // 6. 设置DeepSeek模型
    model, _ := deepseek.NewChatModel(ctx, &deepseek.ChatModelConfig{
        Model:   "deepseek-chat", // 或其他DeepSeek模型
        APIKey:  os.Getenv("DEEPSEEK_API_KEY"),
    })
    
    // 7. 定义RAG提示模板
    ragPrompt := `你是一个基于提供的上下文信息回答问题的助手。
    请仅使用提供的上下文来回答问题，如果上下文中不包含答案，请说"我不知道"。
    
    上下文:
    {{context}}
    
    问题: {{query}}
    
    回答:`
    
    // 8. 构建知识库索引流程
    indexingGraph := compose.NewGraph[document.Source, []string]()
    _ = indexingGraph.AddLoaderNode("loader", loader)
    _ = indexingGraph.AddDocumentTransformerNode("splitter", splitter)
    _ = indexingGraph.AddIndexerNode("indexer", indexer)
    _ = indexingGraph.AddEdge(compose.START, "loader")
    _ = indexingGraph.AddEdge("loader", "splitter")
    _ = indexingGraph.AddEdge("splitter", "indexer")
    _ = indexingGraph.AddEdge("indexer", compose.END)
    
    indexingRunnable, _ := indexingGraph.Compile(ctx, compose.WithGraphName("知识库索引流程"))
    
    // 执行索引流程
    _, err := indexingRunnable.Invoke(ctx, document.Source{
        Path: "./知识库文档路径",
    })
    if err != nil {
        fmt.Printf("索引失败: %v\n", err)
        return
    }
    
    // 9. 构建查询回答流程
    queryGraph := compose.NewGraph[string, *schema.Message]()
    
    // 添加检索节点
    _ = queryGraph.AddLambdaNode("retriever", func(ctx context.Context, query string) (map[string]any, error) {
        docs, err := retriever.Retrieve(ctx, query)
        if err != nil {
            return nil, err
        }
        
        // 合并文档内容作为上下文
        var context string
        for i, doc := range docs {
            context += fmt.Sprintf("%d. %s\n", i+1, doc.Content)
        }
        
        return map[string]any{
            "query":   query,
            "context": context,
        }, nil
    })
    
    // 添加提示模板节点
    template := schema.Jinja2.FromString(ragPrompt)
    _ = queryGraph.AddChatTemplateNode("prompt", template)
    
    // 添加模型节点
    _ = queryGraph.AddChatModelNode("model", model)
    
    // 连接边
    _ = queryGraph.AddEdge(compose.START, "retriever")
    _ = queryGraph.AddEdge("retriever", "prompt")
    _ = queryGraph.AddEdge("prompt", "model")
    _ = queryGraph.AddEdge("model", compose.END)
    
    queryRunnable, _ := queryGraph.Compile(ctx, compose.WithGraphName("RAG查询流程"))
    
    // 10. 执行查询
    userQuery := "Eino框架的主要组件有哪些？"
    response, err := queryRunnable.Invoke(ctx, userQuery)
    if err != nil {
        fmt.Printf("查询失败: %v\n", err)
        return
    }
    
    fmt.Printf("问题: %s\n回答: %s\n", userQuery, response.Content)
}

multiRetriever, err := multiquery.NewRetriever(ctx, &multiquery.Config{
    // 使用DeepSeek模型生成多个查询
    RewriteLLM: deepSeekModel,
    // 原始检索器
    OrigRetriever: basicRetriever,
    // 最大查询数量
    MaxQueriesNum: 3,
})

// 使用多查询检索
docs, err := multiRetriever.Retrieve(ctx, "Eino框架如何构建RAG应用?")

// 创建工具
tools := []*schema.ToolInfo{
    {
        Name:        "search",
        Description: "搜索引擎，用于查找信息",
        Parameters:  schema.NewJSONSchemaProperties().AddString("query", "搜索查询", true),
    },
}

// 创建ReAct智能体
reactAgent, err := react.NewReActAgent(ctx, &react.ReActConfig{
    Model:     deepSeekModel,
    Tools:     tools,
    MaxTurns:  5,
    SystemPrompt: "你是一个有用的助手，能够思考并解决复杂问题。",
})

// 处理用户问题
response, err := reactAgent.Run(ctx, "北京明天的天气如何？")