刘耀文
Home
Posts
Notes
Timeline
Thinking
More
The silent page awaits the waking dawn.
Category
25
articles · from 2025
Agent 与 LLM
2026
22 entries
From RAG to Agent: Engineering Design for Production-Grade AI Systems
#RAG
,
#Agent
+3
Sep 22
How Uber Stabilized AI Costs After 9.4x Growth in Agent Requests
#uber
,
#ai-agent
+6
Aug 31
Why I Built DSHX: Adding a Plugin Development Workflow for DeepSeek Harness
#dshx
,
#deepseek-harness
+6
Aug 27
A Missing Piece in MCP: Skills Over MCP
#mcp
,
#agent-skills
+6
Aug 27
Agent Runtime Series (Part 16): Demo: Implementing a Recoverable, Projectable, Plugin-based Agent Web Runtime
#agent
,
#agent-runtime
+7
Aug 15
Agent Runtime Series (Part 15): DeepSeek Harness's Plugin-Based UI: Slot, Hook, Props, and React
#agent
,
#agent-runtime
+5
Aug 15
Agent Runtime Series (14): Where Does Agent UI State Come From: Event, Projection, and Client Runtime
#agent
,
#agent-runtime
+5
Aug 15
Agent Runtime (Part 13): How to Separate Session, Run, and Connection When Agents Enter the Web
#agent
,
#agent-runtime
+5
Aug 15
Agent Runtime 系列（十二）：Demo：实现一个会随依赖变化自动启停的 Plugin Runtime
#agent
,
#agent-runtime
+5
Aug 15
Agent Runtime Series (11): Cordis and DeepSeek Harness: How the Dependency Chain Re-stabilizes After Plugin Unload
#agent
,
#agent-runtime
+4
Aug 15
Agent Runtime Series (Part 10): Spatiotemporal Composability: How Dynamic Plugin Systems Handle Side Effects and Dependencies
#agent
,
#agent-runtime
+4
Aug 15
Agent Runtime Series (9): Why Agent Core Keeps Getting Heavier: Exploring Harness Extension Boundaries through Pi Extension
#agent
,
#agent-runtime
+4
Aug 15
Agent Runtime Series (6): Pi Session Tree: How a Single JSONL Stores Branches, Compaction, and Current Context
#agent
,
#agent-runtime
+7
Aug 15
Agent Runtime Series (Part 8): Demo — Implementing a Crash-Recoverable Agent Session
#agent
,
#agent-runtime
+8
Aug 15
Agent Runtime Series (7): Why DeepSeek Harness Made Session an Event Log
#agent
,
#agent-runtime
+8
Aug 15
Agent Runtime (Part 5): What Should Be Saved in Agent Sessions: Session, Message, Event, and Context
#agent
,
#agent-runtime
+8
Aug 15
...
#agent
,
#agent-runtime
+8
Aug 14
Agent Runtime Series (Part 3): Pi Agent Loop Source Code: How a Prompt Flows Through Model, Tool, and Next-Round Decision
#agent
,
#agent-runtime
+7
Aug 14
Agent Runtime Series (Part 2): Why Agents Need Multi-Layer Lifecycles: Session, Run, Turn, and Step
#agent
,
#agent-runtime
+7
Aug 14
Agent Runtime Series (1): From Vercel AI SDK, Pi to DeepSeek Harness: The Boundaries of Agent Runtime
#agent
,
#agent-runtime
+5
Aug 14
Writing Experience into the Repository: Some Thoughts on Skills
#AI
,
#Skills
+2
Mar 11
xMemory: Agent Memory Beyond RAG
#AI
,
#智能体
+2
Feb 27
2025
3 entries
构建LLM代理：为什么“多代理”架构往往是坑？
#AI
,
#LLM
+2
Jun 13
模型上下文协议(MCP)深度解析：规范、实践与示例
#AI
,
#MCP
+2
Apr 28
人工智能代理互操作性的未来：谷歌A2A协议与Anthropic MCP的协同作用
#AI
,
#A2A
+2
Apr 14
Tags in this category
Back to top
#agent
18
#agent-runtime
17
#agent-runtime-series
8
#deepseek-harness
8
#pi
7
#plugin-system
6
#session
6
#AI
5
#cordis
4
#event-log
4
#web-runtime
4
#demo
3
#harness
3
#llm
3
#persistence
3
#智能体
3
#LLM
2
#MCP
2
#RAG
2
#context
2
#crash-recovery
2
#follow-up
2
#javascript
2
#jsonl
2
#mcp
2
#projection
2
#react
2
#run
2
#sse
2
#typescript
2
#A2A
1
#Agent
1
#SSE
1
#Skills
1
#agent-loop
1
#agent-skills
1
#ai-agent
1
#client-runtime
1
#compaction
1
#context-engineering
1
#developer-tools
1
#dshx
1
#event
1
#event-sourcing
1
#extension
1
#finops
1
#inbox
1
#inject
1
#input-delivery
1
#lifecycle
1
#mcp-resources
1
#message
1
#plugin-development
1
#prompt-injection
1
#sep-2640
1
#session-tree
1
#skills-over-mcp
1
#slot
1
#software-engineering
1
#spatiotemporal-composability
1
#steer
1
#steering
1
#step
1
#stream-resume
1
#surface
1
#tool-calling
1
#tool-use
1
#turn
1
#uber
1
#vercel-ai-sdk
1
#vite
1
#协议
1
#多智能体
1
#智能体互操作
1
#知识管理
1
#系统设计
1
#自动化
1
#记忆系统
1
Category: Agent 与 LLM - 刘耀文