AI Agent 记忆系统集成方案评估与选型提示词

You are an expert AI systems architect specializing in agent memory infrastructure. I need you to help me evaluate and select the right memory system for my AI agent project.

My Requirements

• Agent framework: [e.g., LangChain / CrewAI / custom Python agent]
• LLM provider: [e.g., OpenAI / Anthropic / local models]
• Scale: [e.g., single agent / multi-agent team / production SaaS]
• Memory needs: [e.g., conversation history / learned facts / cross-session persistence / cross-agent sharing]
• Deployment: [e.g., cloud API / self-hosted / hybrid]

Please Analyze

1. Feature Matrix: Compare Memori, mem0, memsearch, and Letta (MemGPT) across: persistence model, cross-agent memory sharing, structured vs unstructured recall, LLM-agnostic support, SDK languages, and pricing.
2. Architecture Fit: Based on my requirements, recommend the top 2 options with reasoning.
3. Integration Effort: For each recommendation, provide a step-by-step integration plan with estimated effort (hours).
4. Migration Path: If I start with one and need to switch later, what is the data portability story?
5. Production Checklist: Key considerations for taking agent memory to production (latency, cost, privacy, backup).

Output as a structured report with comparison tables and actionable next steps.