update doc about bedrock flow agent & direct call for classifier

docs/src/content/docs/agents/overview.mdx

@@ -180,4 +180,49 @@ When creating a new agent, you can specify various options using the `AgentOptio
   </TabItem>
 </Tabs>
 
+### Direct Agent Usage
+
+When you have a single agent use case, you can bypass the orchestrator and call the agent directly. This approach leverages the power of the Multi Agent Orchestrator framework while focusing on a single agent scenario:
+
+<Tabs syncKey="runtime">
+  <TabItem label="TypeScript" icon="seti:typescript" color="blue">
+    ```typescript
+    // Initialize the agent
+    const agent = new CustomAgent({
+      name: "custom-agent",
+      description: "Handles specific tasks"
+    });
+
+    // Call the agent directly
+    const response = await agent.agentProcessRequest(
+      userInput,
+      userId,
+      sessionId,
+      chatHistory,
+      { param1: "value1" }
+    );
+    ```
+  </TabItem>
+  <TabItem label="Python" icon="seti:python">
+    ```python
+    # Initialize the agent
+    agent = CustomAgent(
+        name="custom-agent",
+        description="Handles specific tasks"
+    )
+
+    # Call the agent directly
+    response = await agent.agent_process_request(
+        input_text=user_input,
+        user_id=user_id,
+        session_id=session_id,
+        chat_history=chat_history,
+        additional_params={"param1": "value1"}
+    )
+    ```
+  </TabItem>
+</Tabs>
+
+This approach is useful for single agent scenarios where you don't need orchestration but want to leverage the powerful capabilities of the Multi Agent Orchestrator framework.
+
 These options allow you to customize various aspects of the agent's behavior and configuration.

docs/src/content/docs/classifiers/overview.mdx

@@ -147,6 +147,50 @@ When deciding between different classifiers, consider:
 
 By thoroughly testing and debugging your chosen Classifier, you can ensure accurate intent classification and efficient query routing in your Multi-Agent Orchestrator.
 
+## Direct Classifier Access
+
+### With Context Management
+Test the classifier with full conversation history handling:
+
+<Tabs syncKey="runtime">
+  <TabItem label="TypeScript" icon="seti:typescript" color="blue">
+    ```typescript
+    const response = await orchestrator.classifyRequest(userInput, userId, sessionId);
+    ```
+  </TabItem>
+  <TabItem label="Python" icon="seti:python">
+    ```python
+    response = await orchestrator.classify_request(user_input, user_id, session_id)
+    ```
+  </TabItem>
+</Tabs>
+
+This method:
+- Retrieves conversation history automatically
+- Maintains context across test calls
+- Ideal for end-to-end testing
+
+### Without Context
+Test raw classification without conversation history:
+
+<Tabs syncKey="runtime">
+  <TabItem label="TypeScript" icon="seti:typescript" color="blue">
+    ```typescript
+    const response = await orchestrator.classifier.classify(userInput, []);
+    ```
+  </TabItem>
+  <TabItem label="Python" icon="seti:python">
+    ```python
+    response = await orchestrator.classifier.classify(user_input, [])
+    ```
+  </TabItem>
+</Tabs>
+
+Best for:
+- Prompt tuning
+- Single-input validation
+- Classification unit tests
+
 ---
 
 For more detailed information on each classifier, refer to the [BedrockClassifier](/multi-agent-orchestrator/classifiers/built-in/bedrock-classifier) and [AnthropicClassifier](/classifiers/built-in/anthropic-classifier) documentation pages.

examples/bedrock-flows/readme.md

@@ -1,30 +1,50 @@
-## BedrockFlowsAgent example
-This example, in Python and Typescript, show how to integrate your Bedrock Flows into the multi-agent orchestrator.
-
-This is the flow we used for our testing.
-
-### tech-agent-flow
-
-In this flow we connected an input node to a prompt node and the output of the prompt is connected to an output node.
-
-
-![tech-agent-flow](./tech-agent-flow.png)
-
-
-The prompt node has 2 inputs:
+# BedrockFlowsAgent Example
+
+This example demonstrates how to use the **[BedrockFlowsAgent](https://awslabs.github.io/multi-agent-orchestrator/agents/built-in/bedrock-flows-agent/)** for direct agent invocation, avoiding the multi-agent orchestration when you only need a single specialized agent.
+
+## Direct Agent Usage
+Call your agent directly using:
+
+Python:
+```python
+response = await orchestrator.agent_process_request(
+    user_input,
+    user_id,
+    session_id,
+    classifier_result
+)
+```
+
+TypeScript:
+```typescript
+const response = await orchestrator.agentProcessRequest(
+    userInput,
+    userId,
+    sessionId,
+    classifierResult
+)
+```
+
+This approach leverages the BedrockFlowsAgent's capabilities:
+- Conversation history management
+- Bedrock Flow integration
+- Custom input/output encoding
+
+### Tech Agent Flow Configuration
+The example flow connects:
+- Input node → Prompt node → Output node
+
+The prompt node accepts:
 - question (current question)
 - history (previous conversation)
 
+![tech-agent-flow](./tech-agent-flow.png)
 ![prompt-node-configuration](./prompt-config.png)
 
+📝 **Note**  
+📅 As of December 2, 2024, Bedrock Flows does not include built-in memory management.
 
-📝 Note
-
-📅 As of December 2, 2024, Bedrock Flows does not include a memory feature to retain previous interactions.
-
-In this example, we demonstrate:
-- 1️⃣ How to integrate your flow into a multi-agent orchestrator.
-- 2️⃣ How to incorporate conversation history into your flow to provide a smoother user experience and generate more accurate results.
-
-🚀 Let's get started!
+See the code samples above for complete implementation details.
 
+---
+*Note: For multi-agent scenarios, add your agents to the orchestrator and use `orchestrator.route_request` (Python) or `orchestrator.routeRequest` (TypeScript) to enable classifier-based routing.*