AI 编码会话自动记忆与上下文注入模板

You are an expert in designing memory systems for AI coding assistants. Your task is to create a comprehensive session memory and context injection framework.

Requirements

Design a system that:

1. Auto-captures key decisions, file changes, architecture patterns, and error resolutions during coding sessions
2. Compresses raw session logs into structured, retrievable knowledge using hierarchical summarization
3. Injects relevant context into future sessions based on semantic similarity and recency
4. Manages memory lifecycle: capture → compress → index → retrieve → inject → prune

Output Format

Provide:

• Memory schema (JSON) with fields: timestamp, category, content, embeddings_key, relevance_score
• Capture rules: what to save (decisions, errors, patterns) vs skip (routine operations)
• Compression pipeline: raw logs → structured summaries → knowledge atoms
• Retrieval strategy: when and how to inject context (session start, on-demand, triggered)
• Pruning policy: TTL, relevance decay, deduplication

Context

Target coding agent: [AGENT_NAME] Project type: [PROJECT_TYPE] Typical session length: [DURATION] Key pain points: [PAIN_POINTS]

Start by analyzing the agent's memory needs, then output the complete framework with implementation examples.