JustVugg/colibri：README 來源編輯指南

JustVugg/colibri 的 README 將專案描述為「Run frontier MoE models on hardware you already own , pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦」。本文只整理倉庫可直接核對的內容，不把 star、Fork 或宣傳語當成品質證明。README 在「README」下寫到：Tiny engine, immense model. Run frontier MoE models , 744B to 2.8T parameters , on consumer and heterogeneous hardware, in pure C with zero engine dependencies, by treating storage, RAM, and VRAM as one inference hierarchy.。這說明的是專案邊界，不是已完成的生產驗證。

從 README 的「Core techniques and measured findings」與相關條目，可以先判斷它是否處理你的實際問題：A JIT for weights. Measured routing heat drives a per-layer LRU, a learned。若需求不同，不應只因專案熱度就採用。本文保留原始專案名、命令與元件名，方便回到一手來源核對。 README 另外列出一項可核對的資訊：One hierarchy, not one memory threshold. VRAM, RAM, and NVMe are placement。這類原文條目可用來設計試跑步驟，但不能取代實際環境測試。

README 將運作方式分散在「README」等段落。可確認的線索包括：> Colibrì is an inference engine you can run today, and an open research > platform. Its primary goal is to pursue inference-side performance across > the entire software/hardware boundary , model formats, memory hierarchy, > storage I/O,。本文不把未寫出的架構、效能或安全邊界補成結論；真正的執行鏈仍要配合目錄、設定檔與版本標籤檢查。

第一次安裝應從 README 指出的入口開始。目前可核對的命令是： $ ./coli chat 如果倉庫沒有命令，本文不會自行編造步驟，而是建議先閱讀「The research mission」，確認系統依賴、預設埠與首次初始化。

日常使用取決於專案文件。README 的「README」段落提到：Colibrì treats VRAM, RAM, and storage as one managed memory hierarchy, and it is deliberately a place to test aggressive systems ideas , so there is no SLA on speed, and a hard guarantee on semantics: experiments must earn their place。設定檔、環境變數、權限與資料目錄只在來源明確時才會記錄；沒有寫出的預設值，應在測試環境驗證並保留回滾副本。 同一部分也提到：I/O is part of the engine. Batched expert unions, overlapped reads and。