Chapter 5: Memory, RAG, and Context

Welcome to Chapter 5: Memory, RAG, and Context. In this part of Mastra Tutorial: TypeScript Framework for AI Agents and Workflows, you will build an intuitive mental model first, then move into concrete implementation details and practical production tradeoffs.

Reliable agents depend on structured context, not ever-growing transcripts.

Context Layers

Layer	Purpose
conversation history	short-term turn continuity
working memory	active task state
semantic recall	long-term retrieval of prior knowledge
RAG context	external knowledge grounding

Best Practices

• summarize stale history into compact state
• keep memory writes explicit and scoped
• validate retrieval quality before response generation

Source References

• Mastra Memory Docs
• Mastra RAG Overview

Summary

You now have a maintainable context strategy for long-lived Mastra systems.

Next: Chapter 6: MCP and Integration Patterns

Source Code Walkthrough

explorations/network-validation-bridge.ts

The ValidationCheck interface in explorations/network-validation-bridge.ts handles a key part of this chapter's functionality:

// ============================================================================

export interface ValidationCheck {
  id: string;
  name: string;
  check: () => Promise<ValidationResult>;
}

export interface ValidationResult {
  success: boolean;
  message: string;
  details?: Record<string, unknown>;
  duration?: number;
}

export interface NetworkValidationConfig {
  /**
   * Array of validation checks to run
   */
  checks: ValidationCheck[];

  /**
   * How to combine check results:
   * - 'all': All checks must pass
   * - 'any': At least one check must pass
   * - 'weighted': Use weights (future)
   */
  strategy: 'all' | 'any';

  /**
   * How validation interacts with LLM completion assessment:
   * - 'verify': LLM says complete AND validation passes

This interface is important because it defines how Mastra Tutorial: TypeScript Framework for AI Agents and Workflows implements the patterns covered in this chapter.

explorations/network-validation-bridge.ts

The ValidationResult interface in explorations/network-validation-bridge.ts handles a key part of this chapter's functionality:

  id: string;
  name: string;
  check: () => Promise<ValidationResult>;
}

export interface ValidationResult {
  success: boolean;
  message: string;
  details?: Record<string, unknown>;
  duration?: number;
}

export interface NetworkValidationConfig {
  /**
   * Array of validation checks to run
   */
  checks: ValidationCheck[];

  /**
   * How to combine check results:
   * - 'all': All checks must pass
   * - 'any': At least one check must pass
   * - 'weighted': Use weights (future)
   */
  strategy: 'all' | 'any';

  /**
   * How validation interacts with LLM completion assessment:
   * - 'verify': LLM says complete AND validation passes
   * - 'override': Only validation matters, ignore LLM
   * - 'llm-fallback': Try validation first, use LLM if no checks configured
   */

This interface is important because it defines how Mastra Tutorial: TypeScript Framework for AI Agents and Workflows implements the patterns covered in this chapter.

explorations/network-validation-bridge.ts

The NetworkValidationConfig interface in explorations/network-validation-bridge.ts handles a key part of this chapter's functionality:

}

export interface NetworkValidationConfig {
  /**
   * Array of validation checks to run
   */
  checks: ValidationCheck[];

  /**
   * How to combine check results:
   * - 'all': All checks must pass
   * - 'any': At least one check must pass
   * - 'weighted': Use weights (future)
   */
  strategy: 'all' | 'any';

  /**
   * How validation interacts with LLM completion assessment:
   * - 'verify': LLM says complete AND validation passes
   * - 'override': Only validation matters, ignore LLM
   * - 'llm-fallback': Try validation first, use LLM if no checks configured
   */
  mode: 'verify' | 'override' | 'llm-fallback';

  /**
   * Maximum time for all validation checks (ms)
   */
  timeout?: number;

  /**
   * Run validation in parallel or sequentially
   */

This interface is important because it defines how Mastra Tutorial: TypeScript Framework for AI Agents and Workflows implements the patterns covered in this chapter.

explorations/network-validation-bridge.ts

The ValidatedNetworkOptions interface in explorations/network-validation-bridge.ts handles a key part of this chapter's functionality:

}

export interface ValidatedNetworkOptions {
  /**
   * Maximum iterations before stopping
   */
  maxIterations: number;

  /**
   * Validation configuration
   */
  validation?: NetworkValidationConfig;

  /**
   * Called after each iteration with validation results
   */
  onIteration?: (result: IterationStatus) => void | Promise<void>;

  /**
   * Thread ID for memory
   */
  threadId?: string;

  /**
   * Resource ID for memory
   */
  resourceId?: string;
}

export interface IterationStatus {
  iteration: number;
  llmSaysComplete: boolean;

This interface is important because it defines how Mastra Tutorial: TypeScript Framework for AI Agents and Workflows implements the patterns covered in this chapter.

How These Components Connect

flowchart TD
    A[ValidationCheck]
    B[ValidationResult]
    C[NetworkValidationConfig]
    D[ValidatedNetworkOptions]
    E[IterationStatus]
    A --> B
    B --> C
    C --> D
    D --> E