Update Dockerfile

Dockerfile

@@ -1,46 +1,81 @@
-# Read the doc: https://huggingface.co/docs/hub/spaces-sdks-docker
-# you will also find guides on how best to write your Dockerfile
-
-FROM  ubuntu:20.04
+FROM ubuntu:20.04
 
+# --- Model URL (Spaces can override this via Build Args) ---
 ARG MODEL_DOWNLOAD_LINK
 ENV MODEL_DOWNLOAD_LINK=${MODEL_DOWNLOAD_LINK:-https://huggingface.co/QuantFactory/MN-Violet-Lotus-12B-GGUF/resolve/main/MN-Violet-Lotus-12B.Q4_K_M.gguf?download=true}
 
+# llama.cpp ref:
+# PR #13249 ("move end-user examples to tools directory") was merged as 1d36b367;
+# we pin to its parent so examples/server/* still exists and your patch applies.
+ARG LLAMA_CPP_REF="1d36b367^"
+
 ENV DEBIAN_FRONTEND=noninteractive
 
+# Spaces runs as UID 1000; create it early and set WORKDIR before COPY
 RUN useradd -m -u 1000 user
-USER user
-ENV PATH="/home/user/.local/bin:$PATH"
-
-WORKDIR /app
+ENV HOME=/home/user
+ENV PATH="$HOME/.local/bin:$PATH"
+WORKDIR $HOME/app
 
-COPY --chown=user . /app
+# Copy your repo (must include helloworld.patch and replace_hw.py at repo root)
+COPY --chown=user . $HOME/app
 
 USER root
 
-RUN apt-get update && apt-get install -y git cmake build-essential g++ wget curl python3
+# System deps
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    git cmake build-essential g++ \
+    wget curl ca-certificates \
+    python3 \
+  && rm -rf /var/lib/apt/lists/*
+
+# Node (needed for the older examples/server/webui build)
+RUN curl -fsSL https://deb.nodesource.com/setup_20.x | bash - \
+  && apt-get update && apt-get install -y --no-install-recommends nodejs \
+  && rm -rf /var/lib/apt/lists/*
+
+# Prepare /data (mounted at runtime if you enable persistent storage on Spaces)
+RUN mkdir -p /data && chmod 777 /data
 
-RUN curl -fsSL https://deb.nodesource.com/setup_18.x | bash -
-RUN apt-get install -y nodejs
-    
 USER user
 
+# Update your patch/UI text based on MODEL_DOWNLOAD_LINK
 RUN python3 replace_hw.py
-RUN git clone https://github.com/ggerganov/llama.cpp.git
 
-WORKDIR /app/llama.cpp
-RUN git apply ../helloworld.patch
+# Clone llama.cpp and pin to a revision compatible with your patch
+RUN git clone https://github.com/ggml-org/llama.cpp.git $HOME/llama.cpp
+WORKDIR $HOME/llama.cpp
+RUN git checkout "${LLAMA_CPP_REF}"
+
+# Apply your UI patch (now updated by replace_hw.py)
+RUN git apply $HOME/app/helloworld.patch
 
-WORKDIR /app/llama.cpp/examples/server/webui
-RUN npm i
+# Build the legacy web UI (exists in this pinned revision)
+WORKDIR $HOME/llama.cpp/examples/server/webui
+RUN npm install
 RUN npm run build
 
-WORKDIR /app/llama.cpp
-RUN cmake -B build -DBUILD_SHARED_LIBS=OFF
-RUN cmake --build build --config Release -j 8
+# Build llama-server (CMake)
+WORKDIR $HOME/llama.cpp
+RUN cmake -B build -DBUILD_SHARED_LIBS=OFF -DCMAKE_BUILD_TYPE=Release
+RUN cmake --build build --config Release -j $(nproc) -t llama-server
 
-WORKDIR /app/llama.cpp/build/bin
-RUN wget -nv -O local_model.gguf "${MODEL_DOWNLOAD_LINK}"
+# Spaces default exposed port is 7860 (set app_port in README if you change it)
+EXPOSE 7860
 
-# HF Docker Spaces default expects your app on port 7860 (or set app_port in README)
-CMD ["/app/llama.cpp/build/bin/llama-server", "--host", "0.0.0.0", "--port", "7860", "-c", "2048", "-m", "local_model.gguf", "--cache-type-k", "q8_0" ]
+# Download model at runtime into /data and start server
+CMD ["bash", "-lc", "set -euo pipefail; \
+  MODEL_FILE=$(python3 -c 'import os,urllib.parse; u=os.environ[\"MODEL_DOWNLOAD_LINK\"]; print(os.path.basename(urllib.parse.urlsplit(u).path))'); \
+  MODEL_DIR=/data/models; MODEL_PATH=\"$MODEL_DIR/$MODEL_FILE\"; \
+  mkdir -p \"$MODEL_DIR\"; \
+  if [ ! -f \"$MODEL_PATH\" ]; then \
+    echo \"Downloading model -> $MODEL_PATH\"; \
+    wget -nv -O \"$MODEL_PATH\" \"$MODEL_DOWNLOAD_LINK\"; \
+  fi; \
+  exec \"$HOME/llama.cpp/build/bin/llama-server\" \
+    --host 0.0.0.0 --port 7860 \
+    -c 2048 \
+    -m \"$MODEL_PATH\" \
+    --cache-type-k q8_0 \
+    --alias \"MN-Violet-Lotus-12B\" \
+"]