42 lines
1.9 KiB
Markdown
42 lines
1.9 KiB
Markdown
---
|
||
title: "Context Substrate"
|
||
type: concept
|
||
tags: [ai-agent, memory, architecture]
|
||
last_updated: 2026-04-23
|
||
---
|
||
|
||
## Definition
|
||
AI Agent 的上下文管理技术路线之一（Camp 2）。维护结构化、人类可读的上下文文件（Markdown、知识图谱、上下文容器），跨会话自然累积增长。
|
||
|
||
## Core Philosophy
|
||
**"What context should the AI work inside?"**（而非 Camp 1 的 "what should the AI remember?"）
|
||
|
||
- Nothing gets extracted — the context is the files.
|
||
- 文件是人类可读、可编辑、可理解的。
|
||
- 因为上下文是文件，人可以随时纠正、补充和理解 Agent 知道什么。
|
||
- 系统随时间自然复合增长（compounding），而非依赖提取质量。
|
||
|
||
## Mechanism
|
||
```
|
||
Agent reads structured context → Agent works within that context →
|
||
Agent (or background process) writes back to the structured context →
|
||
Next session, the context is richer than before
|
||
```
|
||
|
||
## Representative Tools
|
||
- [[OpenClaw]]：Markdown 文件 + dreaming cycle
|
||
- [[Zep]]：Temporal knowledge graph（Graphiti）
|
||
- [[Thoth]]：Personal knowledge graph（10 entity types, 67 relations）
|
||
- [[TrustGraph]]：Context Cores（可移植版本化上下文捆绑包）
|
||
- [[MemSearch]]：Markdown-first + shadow vector index
|
||
- [[ALIVE]]：Structured context substrate, walnuts as portable containers
|
||
|
||
## Relationship to Camp 1
|
||
- Camp 1 优化目标：**召回**（can the system find the right fact?）
|
||
- Camp 2 优化目标：**复合**（does the system get better over time?）
|
||
- Zep 从"memory"→"context engineering"的品牌重塑，是 Camp 1/Camp 2 边界处最强的市场信号
|
||
- Supermemory（Camp 1）的时序感知和 Honcho（Camp 1）的心理建模，代表 Camp 1 向 Camp 2 的演进趋势
|
||
|
||
## Key Distinction from RAG
|
||
RAG 通常指一次性的文档检索问答场景；Context Substrate 强调**跨时间的上下文累积**，是持续运行 Agent 的基础设施。
|