Agent Architecture Patterns

What problem does it solve?

This collection defines actionable architectural patterns for local AI agents, enabling reliable reasoning, structured planning, deterministic tool use, coordinated collaboration, and memory-aware context management.

Core Features & Use Cases

• Reflection pattern to reduce hallucinations via iterative self-critique and refinement.
• Planning pattern to decompose complex goals into atomic tasks with clear validation.
• Tool Use pattern to interact with external environments through a request-response loop.
• Multi-Agent collaboration to assign roles, hand off data, and aggregate outputs.
• Memory & Context pattern to maintain state and context across sessions.

Quick Start

Apply these patterns to a task by drafting an initial agent response, evaluating it critically, and refining it before finalization.