File size: 4,549 Bytes
0c3325c | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 | # ─────────────────────────────────────────────────────────────────────────────
# OpenJev decision API — ZERO GPU (pure CPU)
#
# Official, unmodified openjev-server, pointed at llama.cpp (CPU GGUF) instead
# of vLLM:
#
# 1. builder — compiles llama.cpp `llama-server` for CPU (no CUDA/HIP/Metal)
# and `pip install`s the official openjev-server package
# 2. runtime — runs llama.cpp on CPU + openjev-server on $PORT (7860 on HF)
#
# Why this works: openjev-server's `vllm` backend is really "any OpenAI-
# compatible server that returns logprobs". It sends standard /chat/completions
# JSON (logprobs + top_logprobs + allowed_token_ids) and, when vLLM's
# `logprob_token_ids` extension is absent, falls back to matching the top-K
# logprobs by label text (see openjev_server/backends/vllm.py `_scores`).
# llama.cpp's OpenAI endpoint supports exactly that, so the official code path
# runs unmodified. --force is set because llama.cpp's chat template follows the
# GGUF so closely that the probe letter-mass check stays above the threshold.
#
# HF Spaces ingress :7860 -> openjev-server (UI, /v1/systemone, /docs, ops)
# internal 127.0.0.1:8080 -> llama.cpp (OpenAI-compatible, logprobs on)
# ─────────────────────────────────────────────────────────────────────────────
FROM python:3.12-slim AS builder
WORKDIR /build
ENV DEBIAN_FRONTEND=noninteractive
RUN apt-get update && apt-get install -y --no-install-recommends \
build-essential git cmake ninja-build curl ca-certificates \
&& rm -rf /var/lib/apt/lists/*
# ---- llama.cpp: CPU-only build (no CUDA/HIP/Metal flags anywhere) -----------
ARG LLAMA_CPP_REF=b4630
RUN git clone https://github.com/ggml-org/llama.cpp.git /llama.cpp \
&& cd /llama.cpp \
&& git checkout ${LLAMA_CPP_REF} \
&& git submodule update --init --depth 1 \
&& cmake -B build -DCMAKE_BUILD_TYPE=Release -DGGML_NATIVE=OFF -DGGML_CPU_ALL=ON \
&& cmake --build build --target llama-server -j"$(nproc)" \
&& cmake --install build --prefix /opt/llama
# ---- official openjev-server (unmodified, straight from the upstream repo) --
ARG OPENJEV_SERVER_REF=main
RUN git clone --depth 1 https://github.com/abhishekgahlot2/openjev-server.git /openjev-server \
&& cd /openjev-server \
&& git checkout ${OPENJEV_SERVER_REF} \
&& pip install --no-cache-dir . \
&& cp -r profiles /build/profiles
# ─────────────────────────────────────────────── runtime
FROM python:3.12-slim AS runtime
WORKDIR /app
ENV DEBIAN_FRONTEND=noninteractive \
PYTHONUNBUFFERED=1 \
PIP_NO_CACHE_DIR=1 \
HF_HOME=/data/.cache/huggingface \
OPENJEV_HOST=0.0.0.0 \
OPENJEV_PORT=7860 \
OPENJEV_BACKEND=vllm \
OPENJEV_VLLM_URL=http://127.0.0.1:8080/v1 \
OPENJEV_MODEL=openjev/openjev \
OPENJEV_SERVED_MODEL_NAME=openjev \
OPENJEV_PROFILE=openjev \
OPENJEV_TOKEN="" \
MODEL_REPO=openjev/openjev-GGUF \
MODEL_FILE=OpenJev-Q4_K_M.gguf \
TOKENIZER_REPO=openjev/openjev \
LLAMA_THREADS="" \
LLAMA_THREADS_BATCH="" \
LLAMA_CTX=8192 \
LLAMA_MAX_LOGPROBS=64 \
LLAMA_PARALLEL=1 \
LLAMA_PORT=8080
RUN apt-get update && apt-get install -y --no-install-recommends \
curl ca-certificates libgomp1 \
&& rm -rf /var/lib/apt/lists/* \
&& useradd -m -u 1000 app \
&& mkdir -p /data /app/models \
&& curl -fsSL https://raw.githubusercontent.com/abhishekgahlot2/openjev-server/main/README.md -o /usr/share/OPENJEV-README.md
# llama.cpp runtime binaries
COPY --from=builder /opt/llama /opt/llama
ENV PATH="/opt/llama/bin:${PATH}"
# openjev-server (installed in builder)
COPY --from=builder /usr/local/lib/python3.12/site-packages /usr/local/lib/python3.12/site-packages
COPY --from=builder /usr/local/bin /usr/local/bin
COPY --from=builder /build/profiles /usr/local/lib/python3.12/site-packages/profiles
COPY entrypoint.sh ./
RUN chmod +x /app/entrypoint.sh \
&& chown -R app:app /data /app
USER app
EXPOSE 7860 8080
HEALTHCHECK --interval=30s --timeout=10s --start-period=900s --retries=3 \
CMD curl -fsS http://127.0.0.1:7860/healthz || exit 1
ENTRYPOINT ["./entrypoint.sh"] |