Dev builds FailEcho, a scanner that finds agent failures repeating across runs
EvenAd1183 · reddit · 2026-09-16
A developer noticed repeatedly-running agents (nightly scrapes, scheduled reports, queue workers) keep hitting the same external failures — like a recurring 429 — and small local models silently retry and move on without reporting.
He built FailEcho, MIT-licensed and self-hostable:
- Claude Code version: one stdlib-only file reading local logs, surfacing tool failures repeated across sessions (tool names and counts only, no error text or paths)
- Wrapper version: wraps tools in LangChain/LlamaIndex/custom loops, logging failure shape (service, operation, error class, duration) at 3ms overhead
He's also polling the community: what failure does your pipeline keep hitting, and is it unique to your setup or shared?
Related event: FailEcho Helps Agents Remember Past Errors Instead of Retrying Blindly(2 posts)→
More from coding & agent
- General Reasoning explains how to train agent swarms with RL and latency pressure — rosstaylor90 · 2026-11-02
- "Coding isn't the bottleneck anymore — usage windows are," says founder Kyle Gawley — kylegawley · 2026-11-02
- Demo passed ≠ production-ready: 5 questions to answer before letting agents run your business — _kellhus · 2026-11-02
- An AI agent passed 100 test chats—still not ready to run the business — djgoel123 · 2026-11-02
- Agent given total creative freedom produces a 1:44 code-drawn animated short film — ivegotnoidea1 · 2026-11-02
- Meme: Wake Up, Use Claude Code, Repeat All Day — tekbog · 2026-11-02