Skip to content

README.md

Latest commit

 

History

History
48 lines (30 loc) · 1.64 KB

README.md

File metadata and controls

48 lines (30 loc) · 1.64 KB
Raw

Use Cases

Real scenarios where MisakaNet saves time.

1. Debug Before Asking LLM

Scenario: Your agent hits an error. Instead of asking GPT/Claude (costs tokens, may hallucinate), search MisakaNet first.

python3 search_knowledge.py "chromadb checkpoint error" --top 3

If a lesson exists, you get the fix in 2 seconds. No API call needed.

When to use: Any error that looks like something others have seen. Especially useful for:

  • Infrastructure errors (Docker, SQLite, GitHub Actions)
  • Library-specific bugs (ChromaDB, LangChain, Playwright)
  • Environment issues (WSL, proxy, encoding)

2. Agent Failure Memory

Scenario: Your AI agent (Claude, GPT, Cursor) encounters the same error repeatedly across sessions. Each time, it re-discovers the fix from scratch.

With MisakaNet: The agent searches before debugging. If the fix exists, it skips 10 minutes of trial-and-error.

# Agent wraps its process with fatal-guard
npx @misaka-net/fatal-guard -- node app.js

# On crash: tombstone → draft lesson → bench → verified lesson
# Next agent that hits the same error: instant fix

3. CI/DevOps Incident Reuse

Scenario: Your GitHub Actions workflow fails. The error message is cryptic. You Google it, find a Stack Overflow answer, try 3 approaches, one works.

With MisakaNet: Search the error, get a lesson with the exact fix, verify it works.

python3 search_knowledge.py "GitHub Actions code injection" --domain devops

Bonus: After fixing, write a lesson so the next person (or agent) doesn't repeat the process.

Have a use case to share? Write a field report.