{"id":"iiANgZpNDpEYydBWbLHb8","title":"llama.cpp hsdecide server Dockerfile (template, fork URL needed)","language":"dockerfile","tags":[],"content":"# Dockerfile for a llama.cpp server patched with the hsdecide DecisionType\n# (required to serve nanoandrew4/hs-decide-1-flash-GGUF, which stock llama.cpp can't load)\nFROM ghcr.io/ggml-org/llama.cpp:server-cuda12-b12211 AS base\n# ^ stock image, no hsdecide support -- shown only to pin a base/registry pattern\n\n# Build stage: clone the fork that actually has the hsdecide patch, build the server\nFROM ubuntu:22.04 AS build\nRUN apt-get update && apt-get install -y git cmake build-essential curl libcurl4-openssl-dev\nRUN git clone <URL-OF-THE-HSDECIDE-FORK> /src && cd /src && cmake -B build -DGGML_CUDA=ON && cmake --build build --config Release -j\"$(nproc)\" --target llama-server\n\nFROM ubuntu:22.04\nRUN apt-get update && apt-get install -y libcurl4\nCOPY --from=build /src/build/bin/llama-server /usr/local/bin/llama-server\nENTRYPOINT [\"/usr/local/bin/llama-server\"]\n","created_at":"2026-11-13T11:48:37.585Z"}