Open AIDoc

建立與配置子圖

以下章節提供在建立代理式工作流程 (agentic workflows) 子圖時的程式碼範本與常用模式。

基本子圖建立

自訂子圖通常使用以下模式建立：

具有指定工具選擇策略的子圖：

=== "Kotlin"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.ToolSelectionStrategy
import ai.koog.agents.core.dsl.builder.strategy
import ai.koog.agents.core.dsl.builder.node
import ai.koog.agents.core.dsl.builder.subgraph
typealias StrategyInput = Unit
typealias StrategyOutput = Unit
typealias Input = Unit
typealias Output = Unit
val strategy =
-->
```kotlin
strategy<StrategyInput, StrategyOutput>("strategy-name") {
    val subgraphIdentifier by subgraph<Input, Output>(
        name = "subgraph-name",
        toolSelectionStrategy = ToolSelectionStrategy.ALL
    ) {
        // 為此子圖定義節點與邊
    }

    nodeStart then subgraphIdentifier then nodeFinish
}
```
<!--- KNIT example-custom-subgraphs-01.kt -->

=== "Java"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.AIAgentGraphStrategy;
import ai.koog.agents.core.agent.entity.AIAgentSubgraph;
import ai.koog.agents.core.agent.entity.ToolSelectionStrategy;
class exampleCustomSubgraphsJava01 {
    public static void main(String[] args) {
-->
<!--- SUFFIX
    }
}
-->
```java
var strategyBuilder = AIAgentGraphStrategy.builder("strategy-name")
    .withInput(String.class)
    .withOutput(String.class);

var subgraphIdentifier = AIAgentSubgraph.builder("subgraph-name")
    .withToolSelectionStrategy(ToolSelectionStrategy.ALL.INSTANCE)
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 為此子圖定義節點與邊
    })
    .build();

var strategy = strategyBuilder
    .edge(strategyBuilder.nodeStart, subgraphIdentifier)
    .edge(subgraphIdentifier, strategyBuilder.nodeFinish)
    .build();
```
<!--- KNIT exampleCustomSubgraphsJava01.java -->

具有指定工具清單（來自定義工具註冊表的工具子集）的子圖：

=== "Kotlin"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.ToolSelectionStrategy
import ai.koog.agents.core.dsl.builder.strategy
import ai.koog.agents.core.dsl.builder.node
import ai.koog.agents.core.dsl.builder.subgraph
import ai.koog.agents.ext.tool.AskUser
import ai.koog.agents.ext.tool.SayToUser
typealias StrategyInput = Unit
typealias StrategyOutput = Unit
typealias Input = Unit
typealias Output = Unit
val firstTool = SayToUser
val secondTool = AskUser
val strategy =
-->
```kotlin
strategy<StrategyInput, StrategyOutput>("strategy-name") {
   val subgraphIdentifier by subgraph<Input, Output>(
       name = "subgraph-name",
       tools = listOf(firstTool, secondTool)
   ) {
        // 為此子圖定義節點與邊
    }
}
```
<!--- KNIT example-custom-subgraphs-02.kt -->

=== "Java"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.AIAgentGraphStrategy;
import ai.koog.agents.core.agent.entity.AIAgentSubgraph;
import ai.koog.agents.ext.tool.AskUser;
import ai.koog.agents.ext.tool.SayToUser;
import java.util.List;
class exampleCustomSubgraphsJava02 {
    public static void main(String[] args) {
        var firstTool = SayToUser.INSTANCE;
        var secondTool = AskUser.INSTANCE;
-->
<!--- SUFFIX
    }
}
-->
```java
var strategyBuilder = AIAgentGraphStrategy.builder("strategy-name")
    .withInput(String.class)
    .withOutput(String.class);

var subgraphIdentifier = AIAgentSubgraph.builder("subgraph-name")
    .limitedTools(List.of(firstTool, secondTool))
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 為此子圖定義節點與邊
    })
    .build();

var strategy = strategyBuilder
    .edge(strategyBuilder.nodeStart, subgraphIdentifier)
    .edge(subgraphIdentifier, strategyBuilder.nodeFinish)
    .build();
```
<!--- KNIT exampleCustomSubgraphsJava02.java -->

若要了解更多關於參數與參數值的資訊，請參閱 subgraph API 參考。若要了解更多關於工具的資訊，請參閱工具。

以下程式碼範例顯示了自訂子圖的實際實作：

=== "Kotlin"

<!--- INCLUDE
import ai.koog.agents.core.dsl.builder.strategy
import ai.koog.agents.core.dsl.builder.node
import ai.koog.agents.core.dsl.builder.subgraph
import ai.koog.agents.core.dsl.extension.*
import ai.koog.agents.ext.tool.AskUser
import ai.koog.agents.ext.tool.SayToUser
val firstTool = SayToUser
val secondTool = AskUser
val strategy =
-->
```kotlin
strategy<String, String>("my-strategy") {
   val mySubgraph by subgraph<String, String>(
      tools = listOf(firstTool, secondTool)
   ) {
        // 為此子圖定義節點與邊
        val sendInput by nodeLLMRequest()
        val executeToolCall by nodeExecuteTools()
        val sendToolResult by nodeLLMSendToolResults()

        edge(nodeStart forwardTo sendInput)
        edge(sendInput forwardTo executeToolCall onToolCalls { true })
        edge(executeToolCall forwardTo sendToolResult)
        edge(sendToolResult forwardTo nodeFinish onTextMessage { true })
    }
}
```
<!--- KNIT example-custom-subgraphs-03.kt -->

=== "Java"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.AIAgentEdge;
import ai.koog.agents.core.agent.entity.AIAgentGraphStrategy;
import ai.koog.agents.core.agent.entity.AIAgentNode;
import ai.koog.agents.core.agent.entity.AIAgentSubgraph;
import ai.koog.agents.ext.tool.AskUser;
import ai.koog.agents.ext.tool.SayToUser;
import ai.koog.prompt.message.Message;
import ai.koog.prompt.message.MessagePart;
import java.util.List;
import java.util.stream.Collectors;
class exampleCustomSubgraphsJava03 {
    public static void main(String[] args) {
        var firstTool = SayToUser.INSTANCE;
        var secondTool = AskUser.INSTANCE;
-->
<!--- SUFFIX
    }
}
-->
```java
var strategyBuilder = AIAgentGraphStrategy.builder("my-strategy")
        .withInput(String.class)
        .withOutput(String.class);

var sendInput = AIAgentNode.llmRequest(null);
var executeToolCall = AIAgentNode.executeTools(null);
var sendToolResult = AIAgentNode.llmSendToolResults(null);

var mySubgraph = AIAgentSubgraph.builder()
    .limitedTools(List.of(firstTool, secondTool))
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 為此子圖定義節點與邊
        subgraph
            .edge(AIAgentEdge.builder()
                .from(subgraph.nodeStart)
                .to(sendInput)
                .build()
            )
            .edge(AIAgentEdge.builder()
                .from(sendInput)
                .to(executeToolCall)
                .onToolCalls()
                .build()
            )
            .edge(executeToolCall, sendToolResult)
            .edge(AIAgentEdge.builder()
                .from(sendToolResult)
                .to(subgraph.nodeFinish)
                .onTextMessage()
                .build()
            )
            .build();

    })
    .build();

var strategy = strategyBuilder
    .edge(strategyBuilder.nodeStart, mySubgraph)
    .edge(mySubgraph, strategyBuilder.nodeFinish)
    .build();
```
<!--- KNIT exampleCustomSubgraphsJava03.java -->

在子圖中配置工具

可以透過幾種方式為子圖配置工具：

直接在子圖定義中：

=== "Kotlin"

<!--- INCLUDE
import ai.koog.agents.core.dsl.builder.strategy
import ai.koog.agents.core.dsl.builder.node
import ai.koog.agents.core.dsl.builder.subgraph
import ai.koog.agents.ext.tool.AskUser
val strategy = strategy<String, String>("my-strategy") {
-->
<!--- SUFFIX
}
-->
```kotlin
val mySubgraph by subgraph<String, String>(
   tools = listOf(AskUser)
 ) {
    // 子圖定義
 }
```
<!--- KNIT example-custom-subgraphs-04.kt -->

=== "Java"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.AIAgentSubgraph;
import ai.koog.agents.ext.tool.AskUser;
import java.util.List;
class exampleCustomSubgraphsJava04 {
    public static void main(String[] args) {
-->
<!--- SUFFIX
    }
}
-->
```java
var mySubgraph = AIAgentSubgraph.builder()
    .limitedTools(List.of(AskUser.INSTANCE))
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 子圖定義
    })
    .build();
```
<!--- KNIT exampleCustomSubgraphsJava04.java -->

來自工具註冊表：

=== "Kotlin"

<!--- INCLUDE
import ai.koog.agents.core.dsl.builder.strategy
import ai.koog.agents.core.dsl.builder.node
import ai.koog.agents.core.dsl.builder.subgraph
import ai.koog.agents.core.tools.ToolRegistry
val toolRegistry = ToolRegistry.EMPTY
val strategy = strategy<String, String>("my-strategy") {
-->
<!--- SUFFIX
}
-->
```kotlin
val mySubgraph by subgraph<String, String>(
    tools = listOf(toolRegistry.getTool("AskUser"))
) {
    // 子圖定義
}
```
<!--- KNIT example-custom-subgraphs-05.kt -->

=== "Java"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.AIAgentSubgraph;
import ai.koog.agents.core.tools.ToolRegistry;
import java.util.List;
class exampleCustomSubgraphsJava05 {
    public static void main(String[] args) {
        var toolRegistry = ToolRegistry.builder().build();
-->
<!--- SUFFIX
    }
}
-->
```java
var mySubgraph = AIAgentSubgraph.builder()
    .limitedTools(List.of(toolRegistry.getTool("AskUser")))
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 子圖定義
    })
    .build();
```
<!--- KNIT exampleCustomSubgraphsJava05.java -->

在執行期間動態配置：

=== "Kotlin"

<!--- INCLUDE
import ai.koog.agents.core.dsl.builder.strategy
import ai.koog.agents.core.dsl.builder.node
import ai.koog.agents.core.dsl.builder.subgraph
val strategy = strategy<String, String>("my-strategy") {
    val node by node<Unit, Unit>("node_name") {
-->
<!--- SUFFIX
    }
}
-->
```kotlin
// 建立一組工具
this.llm.writeSession {
    tools = tools.filter { it.name in listOf("first_tool_name", "second_tool_name") }
}
```
<!--- KNIT example-custom-subgraphs-06.kt -->

=== "Java"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.AIAgentNode;
import java.util.List;
import java.util.stream.Collectors;
class exampleCustomSubgraphsJava06 {
    public static void main(String[] args) {
-->
<!--- SUFFIX
    }
}
-->
```java
var node = AIAgentNode.builder("node_name")
    .withInput(String.class)
    .withOutput(String.class)
    .withAction((input, ctx) -> {
        // 建立一組工具
        ctx.getLlm().writeSession(session -> {
            session.setTools(session.getTools().stream()
                .filter(t -> List.of("first_tool_name", "second_tool_name").contains(t.getName()))
                .collect(Collectors.toList()));
            return null;
        });
        return input;
    })
    .build();
```
<!--- KNIT exampleCustomSubgraphsJava06.java -->

進階子圖技術

多部分策略

複雜的工作流程可以分解為多個子圖，每個子圖處理程序中的特定部分：

=== "Kotlin"

<!--- INCLUDE
import ai.koog.agents.core.dsl.builder.strategy
import ai.koog.agents.core.dsl.builder.node
import ai.koog.agents.core.dsl.builder.subgraph
import ai.koog.agents.ext.tool.AskUser
import ai.koog.agents.ext.tool.SayToUser
typealias A = Unit
typealias B = Unit
typealias C = Unit
val firstTool = AskUser
val secondTool = SayToUser
val strategy =
-->
```kotlin
strategy("complex-workflow") {
   val inputProcessing by subgraph<String, A>(
   ) {
      // 處理初始輸入
   }

   val reasoning by subgraph<A, B>(
   ) {
      // 根據處理後的輸入進行推理
   }

   val toolRun by subgraph<B, C>(
      // 來自工具註冊表的選擇性工具子集
      tools = listOf(firstTool, secondTool)
   ) {
      // 根據推理執行工具
   }

   val responseGeneration by subgraph<C, String>(
   ) {
      // 根據工具結果產生回應
   }

   nodeStart then inputProcessing then reasoning then toolRun then responseGeneration then nodeFinish

}
```
<!--- KNIT example-custom-subgraphs-07.kt -->

=== "Java"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.AIAgentGraphStrategy;
import ai.koog.agents.core.agent.entity.AIAgentSubgraph;
import ai.koog.agents.ext.tool.AskUser;
import ai.koog.agents.ext.tool.SayToUser;
import java.util.List;
class exampleCustomSubgraphsJava07 {
    public static void main(String[] args) {
        var firstTool = AskUser.INSTANCE;
        var secondTool = SayToUser.INSTANCE;
-->
<!--- SUFFIX
    }
}
-->
```java
var strategyBuilder = AIAgentGraphStrategy.builder("complex-workflow")
        .withInput(String.class)
        .withOutput(String.class);

var inputProcessing = AIAgentSubgraph.builder()
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 處理初始輸入
    })
    .build();

var reasoning = AIAgentSubgraph.builder()
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 根據處理後的輸入進行推理
    })
    .build();

var toolRun = AIAgentSubgraph.builder()
    // 來自工具註冊表的選擇性工具子集
    .limitedTools(List.of(firstTool, secondTool))
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 根據推理執行工具
    })
    .build();

var responseGeneration = AIAgentSubgraph.builder()
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        // 根據工具結果產生回應
    })
    .build();

var strategy = strategyBuilder
    .edge(strategyBuilder.nodeStart, inputProcessing)
    .edge(inputProcessing, reasoning)
    .edge(reasoning, toolRun)
    .edge(toolRun, responseGeneration)
    .edge(responseGeneration, strategyBuilder.nodeFinish)
    .build();
```
<!--- KNIT exampleCustomSubgraphsJava07.java -->

最佳實務

在使用子圖時，請遵循以下最佳實務：

將複雜的工作流程分解為子圖：每個子圖應具有明確且專一的職責。
僅傳遞必要的上下文：僅傳遞後續子圖正常運作所需的資訊。
記錄子圖相依性：清楚記錄每個子圖對前一個子圖的預期，以及它為後續子圖提供的內容。
隔離測試子圖：在將子圖整合到策略之前，確保每個子圖都能在各種輸入下正確運作。
考慮 Token 使用量：注意 Token 使用量，尤其是在子圖之間傳遞大型歷程記錄時。

疑難排解

工具不可用

如果工具在子圖中不可用：

檢查工具是否已正確註冊在工具註冊表中。

子圖未按定義及預期的順序執行

如果子圖未按定義的順序執行：

檢查策略定義以確保子圖按正確順序排列。
驗證每個子圖是否已將其輸出正確傳遞給下一個子圖。
確保您的子圖與其餘子圖連接，且可從 nodeStart 到達（並可到達 nodeFinish）。請小心使用條件邊，確保它們涵蓋了所有可能的繼續條件，以免在子圖或節點中受阻。

範例

以下範例顯示如何使用子圖在真實場景中建立代理策略。程式碼範例包含三個定義的子圖：researchSubgraph、planSubgraph 和 executeSubgraph，其中每個子圖在助理流程中都有明確且不同的目的。

=== "Kotlin"

<!--- INCLUDE
import ai.koog.agents.core.dsl.builder.strategy
import ai.koog.agents.core.dsl.builder.node
import ai.koog.agents.core.dsl.builder.subgraph
import ai.koog.agents.core.dsl.extension.nodeExecuteTools
import ai.koog.agents.core.dsl.extension.nodeLLMRequest
import ai.koog.agents.core.dsl.extension.nodeLLMSendToolResults
import ai.koog.agents.core.dsl.extension.onTextMessage
import ai.koog.agents.core.dsl.extension.onToolCalls
import ai.koog.agents.core.tools.SimpleTool
import ai.koog.agents.core.tools.ToolDescriptor
import ai.koog.prompt.dsl.prompt
import ai.koog.serialization.typeToken
import kotlinx.serialization.Serializable
class WebSearchTool: SimpleTool<WebSearchTool.Args>(
    argsType = typeToken<Args>(),
    name = "web_search",
    description = "Search on the web"
) {
    @Serializable
    class Args(val query: String)
    override suspend fun execute(args: Args): String {
        return "Searching for ${args.query} on the web..."
    }
}
class DoAction: SimpleTool<DoAction.Args>(
    argsType = typeToken<Args>(),
    name = "do_action",
    description = "Do something"
) {
    @Serializable
    class Args(val action: String)
    override suspend fun execute(args: Args): String {
        return "Doing action..."
    }
}
class DoAnotherAction: SimpleTool<DoAnotherAction.Args>(
    argsType = typeToken<Args>(),
    name = "do_another_action",
    description = "Do something other"
) {
    @Serializable
    class Args(val action: String)
    override suspend fun execute(args: Args): String {
        return "Doing another action..."
    }
}
-->
```kotlin
// 定義代理策略
val strategy = strategy<String, String>("assistant") {

    // 包含工具呼叫的子圖
    val researchSubgraph by subgraph<String, String>(
        "research_subgraph",
        tools = listOf(WebSearchTool())
    ) {
        val nodeCallLLM by nodeLLMRequest("call_llm")
        val nodeExecuteTool by nodeExecuteTools()
        val nodeSendToolResult by nodeLLMSendToolResults()

        edge(nodeStart forwardTo nodeCallLLM)
        edge(nodeCallLLM forwardTo nodeExecuteTool onToolCalls { true })
        edge(nodeExecuteTool forwardTo nodeSendToolResult)
        edge(nodeSendToolResult forwardTo nodeExecuteTool onToolCalls { true })
        edge(nodeCallLLM forwardTo nodeFinish onTextMessage { true })
    }

    val planSubgraph by subgraph(
        "plan_subgraph",
        tools = listOf()
    ) {
        val nodeUpdatePrompt by node<String, Unit> { research ->
            llm.writeSession {
                rewritePrompt {
                    prompt("research_prompt") {
                        system(
                            "You are given a problem and some research on how it can be solved." +
                                    "Make step by step a plan on how to solve given task."
                        )
                        user("Research: $research")
                    }
                }
            }
        }
        val nodeCallLLM by nodeLLMRequest("call_llm")

        edge(nodeStart forwardTo nodeUpdatePrompt)
        edge(nodeUpdatePrompt forwardTo nodeCallLLM transformed { "Task: $agentInput" })
        edge(nodeCallLLM forwardTo nodeFinish onTextMessage { true })
    }

    val executeSubgraph by subgraph<String, String>(
        "execute_subgraph",
        tools = listOf(DoAction(), DoAnotherAction()),
    ) {
        val nodeUpdatePrompt by node<String, Unit> { plan ->
            llm.writeSession {
                rewritePrompt {
                    prompt("execute_prompt") {
                        system(
                            "You are given a task and detailed plan how to execute it." +
                                    "Perform execution by calling relevant tools."
                        )
                        user("Execute: $plan")
                        user("Plan: $plan")
                    }
                }
            }
        }
        val nodeCallLLM by nodeLLMRequest("call_llm")
        val nodeExecuteTool by nodeExecuteTools()
        val nodeSendToolResult by nodeLLMSendToolResults()

        edge(nodeStart forwardTo nodeUpdatePrompt)
        edge(nodeUpdatePrompt forwardTo nodeCallLLM transformed { "Task: $agentInput" })
        edge(nodeCallLLM forwardTo nodeExecuteTool onToolCalls { true })
        edge(nodeExecuteTool forwardTo nodeSendToolResult)
        edge(nodeSendToolResult forwardTo nodeExecuteTool onToolCalls { true })
        edge(nodeCallLLM forwardTo nodeFinish onTextMessage { true })
    }

    nodeStart then researchSubgraph then planSubgraph then executeSubgraph then nodeFinish
}
```
<!--- KNIT example-custom-subgraphs-08.kt -->

=== "Java"

<!--- INCLUDE
import ai.koog.agents.core.agent.entity.AIAgentEdge;
import ai.koog.agents.core.agent.entity.AIAgentGraphStrategy;
import ai.koog.agents.core.agent.entity.AIAgentNode;
import ai.koog.agents.core.agent.entity.AIAgentSubgraph;
import ai.koog.agents.core.tools.annotations.LLMDescription;
import ai.koog.agents.core.tools.annotations.Tool;
import ai.koog.agents.core.tools.reflect.ToolSet;
import ai.koog.prompt.Prompt;
import ai.koog.prompt.message.Message;
import ai.koog.prompt.message.MessagePart;
import java.util.Collections;
import java.util.stream.Collectors;
public class exampleCustomSubgraphsJava08 {
    static class WebSearchToolSet implements ToolSet {
        @Tool
        @LLMDescription("Search on the web")
        public String webSearch(@LLMDescription("The search query") String query) {
            return "Searching for " + query + " on the web...";
        }
    }
    static class ActionToolSet implements ToolSet {
        @Tool
        @LLMDescription("Do something")
        public String doAction(@LLMDescription("The action to perform") String action) {
            return "Doing action...";
        }
        @Tool
        @LLMDescription("Do something other")
        public String doAnotherAction(@LLMDescription("The action to perform") String action) {
            return "Doing another action...";
        }
    }
    public static void main(String[] args) {
-->
<!--- SUFFIX
    }
}
-->
```java
// 定義代理策略
var strategyBuilder = AIAgentGraphStrategy.builder("assistant")
    .withInput(String.class)
    .withOutput(String.class);

// 包含工具呼叫的子圖
var nodeCallLLM = AIAgentNode.llmRequest(null);
var nodeExecuteTool = AIAgentNode.executeTools(null);
var nodeSendToolResult = AIAgentNode.llmSendToolResults(null);

var researchSubgraph = AIAgentSubgraph.builder("research_subgraph")
    .limitedTools(new WebSearchToolSet())
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        subgraph
            .edge(AIAgentEdge.builder()
                .from(subgraph.nodeStart)
                .to(nodeCallLLM)
                .build()
            )
            .edge(AIAgentEdge.builder()
                .from(nodeCallLLM)
                .to(nodeExecuteTool)
                .onToolCalls()
                .build()
            )
            .edge(nodeExecuteTool, nodeSendToolResult)
            .edge(AIAgentEdge.builder()
                .from(nodeSendToolResult)
                .to(nodeExecuteTool)
                .onToolCalls()
                .build()
            )
            .edge(AIAgentEdge.builder()
                .from(nodeCallLLM)
                .to(subgraph.nodeFinish)
                .onTextMessage()
                .build()
            )
            .build();
    })
    .build();

var nodeUpdatePrompt = AIAgentNode.builder()
    .withInput(String.class)
    .withOutput(String.class)
    .withAction((research, ctx) -> {
        ctx.getLlm().writeSession(session -> {
            session.setPrompt(Prompt.builder("research_prompt")
                .system(
                    "You are given a problem and some research on how it can be solved." +
                    "Make step by step a plan on how to solve given task."
                )
                .user("Research: " + research)
                .build());
            return null;
        });
        return "Task: " + ctx.getAgentInput();
    })
    .build();
var nodeCallLLMPlan = AIAgentNode.llmRequest(null);

var planSubgraph = AIAgentSubgraph.builder("plan_subgraph")
    .limitedTools(Collections.emptyList())
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        subgraph
            .edge(subgraph.nodeStart, nodeUpdatePrompt)
            .edge(AIAgentEdge.builder()
                .from(nodeUpdatePrompt)
                .to(nodeCallLLMPlan)
                .build()
            )
            .edge(AIAgentEdge.builder()
                .from(nodeCallLLMPlan)
                .to(subgraph.nodeFinish)
                .onTextMessage()
                .build()
            )
            .build();
    })
    .build();

var nodeUpdatePromptExecute = AIAgentNode.builder()
    .withInput(String.class)
    .withOutput(String.class)
    .withAction((plan, ctx) -> {
        ctx.getLlm().writeSession(session -> {
            session.setPrompt(Prompt.builder("execute_prompt")
                .system(
                    "You are given a task and detailed plan how to execute it." +
                    "Perform execution by calling relevant tools."
                )
                .user("Execute: " + plan)
                .user("Plan: " + plan)
                .build());
            return null;
        });
        return "Task: " + ctx.getAgentInput();
    })
    .build();

var nodeCallLLMExecute = AIAgentNode.llmRequest(null);
var nodeExecuteToolExecute = AIAgentNode.executeTools(null);
var nodeSendToolResultExecute = AIAgentNode.llmSendToolResults(null);

var executeSubgraph = AIAgentSubgraph.builder("execute_subgraph")
    .limitedTools(new ActionToolSet())
    .withInput(String.class)
    .withOutput(String.class)
    .define(subgraph -> {
        subgraph
            .edge(subgraph.nodeStart, nodeUpdatePromptExecute)
            .edge(AIAgentEdge.builder()
                .from(nodeUpdatePromptExecute)
                .to(nodeCallLLMExecute)
                .build()
            )
            .edge(AIAgentEdge.builder()
                .from(nodeCallLLMExecute)
                .to(nodeExecuteToolExecute)
                .onToolCalls()
                .build()
            )
            .edge(nodeExecuteToolExecute, nodeSendToolResultExecute)
            .edge(AIAgentEdge.builder()
                .from(nodeSendToolResultExecute)
                .to(nodeExecuteToolExecute)
                .onToolCalls()
                .build()
            )
            .edge(AIAgentEdge.builder()
                .from(nodeCallLLMExecute)
                .to(subgraph.nodeFinish)
                .onIsInstance(Message.Assistant.class)
                .onTextMessage()
                .build()
            )
            .build();
    })
    .build();

var strategy = strategyBuilder
    .edge(strategyBuilder.nodeStart, researchSubgraph)
    .edge(researchSubgraph, planSubgraph)
    .edge(planSubgraph, executeSubgraph)
    .edge(executeSubgraph, strategyBuilder.nodeFinish)
    .build();
```
<!--- KNIT exampleCustomSubgraphsJava08.java -->

規劃代理

建立提示

執行提示

聊天記憶體

OpenTelemetry

子圖

建立與配置子圖

基本子圖建立

在子圖中配置工具

進階子圖技術

多部分策略

最佳實務

疑難排解

工具不可用

子圖未按定義及預期的順序執行

範例

建立與配置子圖 ​

基本子圖建立 ​

在子圖中配置工具 ​

進階子圖技術 ​

多部分策略 ​

最佳實務 ​

疑難排解 ​

工具不可用 ​

子圖未按定義及預期的順序執行 ​

範例 ​

建立與配置子圖

基本子圖建立

在子圖中配置工具

進階子圖技術

多部分策略

最佳實務

疑難排解

工具不可用

子圖未按定義及預期的順序執行

範例