This enhanced prompt is designed to turn your request into a structural architectural blueprint. You can paste this directly into an LLM to receive a professional, system-level design document.

The Prompt

Role: You are a Senior AI Systems Architect specializing in Cognitive Architectures and Agentic Memory Systems. Your expertise lies in designing scalable, low-latency memory pipelines for LLM-based autonomous agents.

Context: I am developing a sophisticated multi-turn AI agent that requires a robust, modular memory management system. I need a comprehensive design document that outlines how the agent processes, stores, and retrieves information across different temporal and semantic layers.

Task: Please architect a memory management system for an agent defined by the following [AGENT_PURPOSE]. The output must detail the implementation strategy for the five memory domains listed below.

Memory Domains to Address:

1. Short-term Memory: Define the strategy for managing active conversation history (e.g., sliding window vs. token-budgeting).
2. Long-term Memory: Propose a vector-based retrieval architecture (e.g., RAG, semantic chunking, embedding models).
3. Entity Memory: Describe how to extract, persist, and update a "Knowledge Graph" or "User Profile" database of facts.
4. Importance/Weighting Knobs: Define the logic for "Memory Decay" or "Salience Scoring"—how the system decides what is worth keeping versus discarding.
5. Visual Graph Logic: Propose a schema for visualizing memory nodes and their associations, suitable for debugging and interpretability.

Constraints:

• Tone: Technical, precise, and professional.
• Avoid: Vague terminology; provide concrete frameworks (e.g., mention specific architectures like MemGPT, LangGraph, or Pinecone/Weaviate concepts).
• Length: Keep the response concise but deep—focus on actionable architecture over theoretical fluff.

📝 Output Format

Use the following structure for your response:

• System Overview: A high-level summary of the proposed memory architecture.
• Component Deep-Dive: A section for each of the 5 memory domains, detailing data flow and storage logic.
• Integration Strategy: A brief explanation of how these layers interact during a single inference cycle.
• Implementation Schema: A pseudo-code snippet or JSON structure representing the memory state object.

🧩 Variables

• [AGENT_PURPOSE]: [Insert your agent's goal here, e.g., "A personal executive assistant for software engineers"]
• [SCALE_REQUIREMENTS]: [Insert scale, e.g., "High-volume, thousands of concurrent sessions"]

How to use this:

2. Replace the bracketed variables at the bottom with your specific needs.
3. Execute the prompt to receive a professional-grade technical specification for your AI agent's memory architecture.