Relay: an incident response agent that turns resolved incidents into reusable memory
Top1Coder · reddit · 2026-11-07
A developer shares Relay, an incident response agent built around a memory architecture:
- Hindsight persistent memory: after resolution, retain confirmed root causes, fix steps, and lessons
- Cross-incident recall: retrieve prior experiences even when wording differs
- Local retrieval: TF-IDF + cosine similarity as an inspectable extra path
- LLM reasoning: Gemini combines the current incident, retrieved memories, and runbooks into structured hypotheses
- Human review: retrieved memories are evidence, not truth; engineers approve resolutions
Key lessons: memory quality depends on what you retain (confirmed resolutions, not every generated response), and memory and reasoning should be separate components so the LLM never rediscovers operational history from scratch.
Related event: Dev builds Relay, an incident-response agent with persistent memory(2 posts)→
More from coding & agent
- Orosz: SDKs without a 'prompt for your agent' onboarding won't see adoption — HankYeomans · 2026-11-08
- Git Worktrees: The Safe Way to Run Multiple AI Coding Agents in One Repo — draginol · 2026-11-08
- Security vet Chuvakin writes new blog post largely with his own Gemini skill — AccBalanced · 2026-11-08
- Dev open-sources local-reader: Kokoro TTS reads articles aloud on Mac and iPhone — ethanniser · 2026-11-08
- Open-source app turns any GitHub repo into architecture diagrams and wikis using Kimi K3.1 Pro — Arindam_1729 · 2026-11-08
- Claude is "insane" at Instagram edits but still can't do good sound design — ordax · 2026-11-08