langchain.js
    Preparing search index...

    Function piiRedactionMiddleware

    • Creates a middleware that detects and redacts personally identifiable information (PII) from messages before they are sent to model providers, and restores original values in model responses for tool execution.

      The middleware intercepts agent execution at two points:

      • Applies regex-based pattern matching to all message content (HumanMessage, ToolMessage, SystemMessage, AIMessage)
      • Processes both message text and AIMessage tool call arguments
      • Each matched pattern generates:
        • Unique identifier: generateRedactionId() → "abc123"
        • Redaction marker: [REDACTED_{RULE_NAME}_{ID}] → "[REDACTED_SSN_abc123]"
        • Redaction map entry: { "abc123": "123-45-6789" }
      • Returns modified request with redacted message content
      • Scans AIMessage responses for redaction markers matching pattern: /\[REDACTED_[A-Z_]+_(\w+)\]/g
      • Replaces markers with original values from redaction map
      • Handles both standard responses and structured output (via tool calls or JSON content)
      • For structured output, restores values in both the tool call arguments and the structuredResponse state field
      • Returns new message instances via RemoveMessage/AIMessage to update state
      User Input: "My SSN is 123-45-6789"
          ↓ [beforeModel]
      Model Request: "My SSN is [REDACTED_SSN_abc123]"
          ↓ [model invocation]
      Model Response: tool_call({ "ssn": "[REDACTED_SSN_abc123]" })
          ↓ [afterModel]
      Tool Execution: tool({ "ssn": "123-45-6789" })

      This middleware provides model provider isolation only. PII may still be present in:

      • LangGraph state checkpoints (memory, databases)
      • Network traffic between client and application server
      • Application logs and trace data
      • Tool execution arguments and responses
      • Final agent output

      For comprehensive PII protection, implement additional controls at the application, network, and storage layers.

      Parameters

      • options: InferInteropZodInput = {}

        Configuration options

        • rules

          Record of detection rules mapping rule names to regex patterns. Rule names are normalized to uppercase and used in redaction markers. Patterns must use the global flag (/pattern/g) to match all occurrences.

      Returns AgentMiddleware<any>

      Middleware instance for use with createAgent

      import { piiRedactionMiddleware } from "langchain";
      import { createAgent } from "langchain";
      import { tool } from "@langchain/core/tools";
      import { z } from "zod";

      const PII_RULES = {
        ssn: /\b\d{3}-?\d{2}-?\d{4}\b/g,
        email: /\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}\b/g,
        phone: /\b\d{3}[-.]?\d{3}[-.]?\d{4}\b/g,
      };

      const lookupUser = tool(async ({ ssn }) => {
        // Receives original value: "123-45-6789"
        return { name: "John Doe", account: "active" };
      }, {
        name: "lookup_user",
        description: "Look up user by SSN",
        schema: z.object({ ssn: z.string() })
      });

      const agent = createAgent({
        model: new ChatOpenAI({ model: "gpt-4" }),
        tools: [lookupUser],
        middleware: [piiRedactionMiddleware({ rules: PII_RULES })]
      });

      const result = await agent.invoke({
        messages: [new HumanMessage("Look up SSN 123-45-6789")]
      });
      // Model request: "Look up SSN [REDACTED_SSN_abc123]"
      // Model response: tool_call({ "ssn": "[REDACTED_SSN_abc123]" })
      // Tool receives: { "ssn": "123-45-6789" }

      const agent = createAgent({
        model: new ChatOpenAI({ model: "gpt-4" }),
        tools: [someTool],
        middleware: [piiRedactionMiddleware()]
      });

      // Configure rules at runtime via middleware context
      const result = await agent.invoke(
        { messages: [new HumanMessage("...")] },
        {
          configurable: {
            PIIRedactionMiddleware: {
              rules: {
                ssn: /\b\d{3}-?\d{2}-?\d{4}\b/g,
                email: /\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]{2,}\b/g,
              }
            }
          }
        }
      );

      const customRules = {
        employee_id: /EMP-\d{6}/g,
        api_key: /sk-[a-zA-Z0-9]{32}/g,
        credit_card: /\b\d{4}[- ]?\d{4}[- ]?\d{4}[- ]?\d{4}\b/g,
      };

      const middleware = piiRedactionMiddleware({ rules: customRules });
      // Generates markers like: [REDACTED_EMPLOYEE_ID_xyz789]

    Settings

    Member Visibility

    On This Page

    MechanismData FlowLimitations