MorrowCache: local proxy dedupes paraphrased LLM queries, ~6.7x faster hits
TurnoverSea9119 · reddit · 2026-09-26
A dev open-sourced MorrowCache, a local OpenAI-compatible proxy that cuts duplicate LLM spend. An adjudicator model checks whether an incoming prompt matches a cached answer's intent: HIT reuses the reply and skips the upstream call, MISS calls the model once and stores it, and a flaky judge fails open so requests never get stuck. Judges: cloud Jev (default) or local Kev/Laya/System One — real intent matching, not cosine embedding vibes. Recorded run: a paraphrase went from a 2817ms miss to a 423ms hit, 6.7x faster and mostly free.
More from coding & agent
- Making GA4 Agent-Friendly: Building Analytics Tools for an MCP Server — DutchSEOnerd · 2026-09-26
- Loop engineering beats prompt engineering: build agent loops that self-verify — goyalshaliniuk · 2026-09-26
- "Cognitive Feudalism": a 60-Second Visual Essay Written Frame-by-Frame in Code by Claude — CurieuxExplorer · 2026-09-26
- One Prompt, 1,850 Lines of Code: Claude Builds a Mumbai Walkthrough With No Assets — CurieuxExplorer · 2026-09-26
- Grok Bot Was Hidden in Cursor All Along: Enable the Sidebar for Seamless Handoff — CurieuxExplorer · 2026-09-26
- One prompt, 86 minutes, $25.66: Claude Opus 5.5 builds an interactive lens physics lab — CurieuxExplorer · 2026-09-26