Skip to content
ETF Agent / Notes
Search
K
Main Navigation
All articles
About
English
简体中文
English
简体中文
Toggle Appearance
Directory
All articles
Article categories
Module Deep Dives
L1 · Task Input and Triggers
Scheduled Research Beyond a Conversation
Daily Personalized Research Is Not Just Repeating a Prompt
L2 · Identity, Sessions and Context
Keep Chatting Is Not Keep Executing: Agent Conversation Admission and Task Continuation
How Does an Agent Keep Each User’s Research Private?
Models Understand; Code Authorizes: Run-Scoped Authorization Through Recovery
Extension Host: From Feature Phone to Smartphone
Main Context Compaction Without Rewriting Facts
Cross-Run Compaction: Why an Agent Cannot Simply Summarize History
Why Agents Need a Workspace: From One-Off Conversations to Long-Term Research
Memory: From Saving and Recall to Safe Forgetting
Prompt Injection Is More Than Malicious Text: An Agent Content Trust Boundary
Why an Agent Cannot Execute a Stop Command Immediately
After an Agent Asks: Durable Human Response Adoption
A Custom Expert Should Be More Than a Prompt Field
Agent Usage: Requests, Tokens and Shadow Credits
How Is a Skill Different from a Prompt?
Action Reminders: From Candidate Advice to the Model Request
Related reading · canonical entry elsewhere
Why a Research Report Must Become an Asset Before an Agent Can Reuse It
Externalized Results and Input Completeness
L3 · Agent Decisions and Execution
Entry: Why the Entry Point Became Simpler
When Should an Agent Stop Researching and Start Answering?
Provider: Why a Unified API Still Needs Shared Admission
Why the Next Model Request Still Queues After HTTP 200
Tool Surface: Why the Model Cannot See an Available Tool
Multi-Ecosystem Hooks: Claude, Pi and Codex
Checkpoint System Design: What Does an Agent Save, and What Lets It Continue?
What a Model Attempt Leaves Behind: Capturing, Retaining, and Reading Agent Research Evidence
Related reading · canonical entry elsewhere
Agent Observability Panorama: From Model Requests to Verifiable Research Facts
Deep Research: From Budget Upgrades to Delivery—What We Built and What the Tests Show
Adaptive Research and Python
Subagent Panorama: Independent Context, Specialist Work and Real Handoffs
Run Budget and Constraint Governance
Action Reminders: From Candidate Advice to the Model Request
L4 · Tools and Domain Capabilities
Web Research: From Search to Deep Reading
PDF: From Text to Evidence
When an Agent Says Today: User Time and Market Sessions
Python: From Tool to Computational Workbench
Exact Numbers Are Not Correct Investments: A Numeric Kernel for Units, Rounding, and Residuals
Deterministic Fund Allocation: From Holdings to Reproducible Amount Advice
Why Can’t a Portfolio Report Be Written While It Is Still Being Calculated?
Connecting Portfolio Data Through a Read-Only MCP
The RAG Lifecycle: From Document Ingestion to Cited Answers
MCP Host: From Discovery to Usable Tools
Why a Research Report Must Become an Asset Before an Agent Can Reuse It
Externalized Results and Input Completeness
Related reading · canonical entry elsewhere
Observing the Tool-to-Model Input Lifecycle
RAG: Revocation and Historical Replay
Controlled Refresh for Stale Data
How Is a Skill Different from a Prompt?
What a Model Attempt Leaves Behind: Capturing, Retaining, and Reading Agent Research Evidence
L5 · Result Publication and Delivery
Publication: Why Answers Need Their Own Delivery System
A Report Is Generated: Why Can’t It Be Shared Yet?
When a User Dislikes an Answer: Feedback Is Not Automatic Learning
Related reading · canonical entry elsewhere
Async Research: Delivering on Time After Users Leave
L6 · User Interface
An Agent Client Is More Than an SSE Renderer: Presenting Research Progress Truthfully
Related reading · canonical entry elsewhere
Keeping Progressive Answers Coherent
Architecture and Systems Thinking
Deep and Continuous Research
Agent Client Interaction Panorama: From Asking a Question to Understanding Results and Continuing Research
Long-Horizon Research Panorama: Continuing Across Context, Interruptions and Time
Designing Half-Hour Agent Research: From Ordinary Research to Reliable Delivery
Keep Chatting: Adding Async Background Research to an Agent in a Day
From Investment-Research Chat to Ongoing Research: Lessons from DeepSeek Super-agent
After the Weekly Report, Does the Agent Remember Why?
Why Saved HTML Can Still Fail Delivery
Failover: Change the Route Without Resetting the Task
A Disconnected Browser Is Not a Crashed Agent
Run Termination, Stage Budgets and Safe Recovery
A Research Agent Needs More Than Tools: Model, Harness, and Loop
Deep Research: From Budget Upgrades to Delivery—What We Built and What the Tests Show
Adaptive Research and Python
Async Research: Delivering on Time After Users Leave
Keeping Progressive Answers Coherent
Subagents and Experts
More Than One Investment Expert: How Multi-Perspective Agents Test an Investment Thesis
When Two Agents Talk Directly: Bounded Bull–Bear Research
Subagent Panorama: Independent Context, Specialist Work and Real Handoffs
Related reading · canonical entry elsewhere
How Do We Know the Main Agent Actually Used a Specialist’s Findings?
A Custom Expert Should Be More Than a Prompt Field
Why Did One Subagent Change Hold Up Our Work for an Entire Day?
Financial Research and Business Accountability
After General-Purpose Agent Frameworks, Why Build Financial Domain Capabilities?
From a General Harness to Financial Research: A Four-Way Architecture Comparison of ETF Agent
Financial Agents in Context: Systems, Lifecycles, and Accountability
Trusted Handoffs Between Financial Business Agents: Why a Router Is Not Enough
Why Financial Market MCP Became a Decision System
Boundaries Between Research, Advice and Trading
Financial Data: Time, Definitions and Reproducibility
Why Python Results Can Still Be Misused in Reports
Why Can a Fund Research Report Still Make Unsupported Conclusions?
Architecture Learning Path
Agents in 15 Minutes
Prompt, Context, RAG and Memory — Where Does the Model Input Come From?
Agent Intermediate: Ten Core Modules in Open ETF Agent
Tools, Function Calling, MCP and Skills — How Does an Action Happen?
History, Context, Memory and Checkpoints — How Does an Agent Continue?
Agent Execution Identity Panorama: From a Conversation to Model Turns—What Actually Continues?
Agent Deep Dive: From One Report to Durable Execution and Specialist Collaboration
Bring Your Own Documents to an Agent: From the First Report to a Resumable Research Project
Expert, Subagent, Review, Judge: If They All Use Another Model, Why Are They Different?
After the report: source checks, specialist work, and independent review
Six-Layer Request Lifecycle: From Prompt to Visible Answer
Agent Context Governance Panorama · Principles: Keep Every Decision Aligned with the Goal
Agent Context Governance Panorama · Practice: From the First Request to Long-Term Memory
System Boundaries and Evolution
ETF Agent vs LangGraph: Concept Mapping, Recovery, and Research Delivery
From Chat to Copilot to Agent: What Actually Changes in the Architecture?
Why Split an Agent into Three Processes? Independent Lifecycles Without Splitting Business Truth
How Agent Development Differs from Backend Development: From One Request to a Long-Term Personal Assistant
Nine Architecture Judgments Behind One Report
What to Trust in Agent Engineering: From Local Facts to System Commitments
Agent Fragility and Antifragility: When Safeguards Start Causing Failures
The Failure Is Already Known. Why Hasn't the Call Returned? Cleanup and Result Ownership
Don't Turn the Model into a State Machine
Responsibilities of Model, Runtime, MCP and Skills
A Trustworthy Evidence Chain for Financial Agents
Why AI Agents Need Thinking, Fast and Slow
Designing Agent Citations: From Host-Issued References to Revision-Safe Independent Review
Code Is Cheaper; Why Are Systems Harder to Evolve?
What Should Stay Stable—and What Should Change—in an Agent Architecture?
Stop Blaming Every Failure on the Model
Local Improvements Are Not System Progress
Agent Observability Panorama: From Model Requests to Verifiable Research Facts
What Does an Agent-Generated Report Actually Cost?
Observing the Tool-to-Model Input Lifecycle
AI Agent Engineering Practice
Specs, Evidence and Engineering Memory
AI Coding Quality Panorama: Taking Responsibility for the Whole Project When AI Writes Most of the Code
Agent Engineering Suite: Connecting Requirements to Release
When Tests Outgrow the Code: Helping Docs, Code, and Tests Evolve Together
Does the Code Really Support the Article? An Evidence Contract for Technical Writing
Spec-Driven Agent Development: From Requirements to Trustworthy Delivery
Evidence and Trustworthy Acceptance in Agent Development
Why We Split a Large Weekly Review Spec into Three Phases
Engineering Knowledge: Context, Memory and Retirement
Real Validation and Task Convergence
Start with a Report You Can Read Back: Open ETF Agent v0.1
Agent Evaluation Panorama: From Green Tests to Real Task Value
Agent Evaluation Identity: Can Results Survive Model Failover?
Building Agent Evaluation Sets Across Development and Acceptance
How Do We Know the Main Agent Actually Used a Specialist’s Findings?
How a null Broke an Agent Tool Call: A Streaming Compatibility Case
RAG Recall Looks Great. Why Doesn’t That Prove the Agent Found the Right Financial Evidence?
Why Did One Subagent Change Hold Up Our Work for an Entire Day?
From Bug-Fix Loops to Task Convergence
Why a Small ETF Investment Assistant Became an Engineering Problem
The Expensive Part of AI Coding Is Deciding What Deserves Verification
RAG: Revocation and Historical Replay
Controlled Refresh for Stale Data
Run Budget and Constraint Governance
Tag topics
Finance
Open source · Open ETF Agent
Panorama
More pages
Architecture Gallery
Financial Research Backlog
Blog
↗
Runnable code
↗
中文
/
EN
GitHub
Star
↗
Back to Top
Browse all articles