tokenfactory/ai-station

1# syntax=docker/dockerfile:12# Initialize device type args3# use build args in the docker build command with --build-arg="BUILDARG=true"4ARG USE_CUDA=false5ARG USE_OLLAMA=false6# Tested with cu117 for CUDA 11 and cu121 for CUDA 12 (default)7ARG USE_CUDA_VER=cu1218# any sentence transformer model; models to use can be found at https://huggingface.co/models?library=sentence-transformers9# Leaderboard: https://huggingface.co/spaces/mteb/leaderboard 10# for better performance and multilangauge support use "intfloat/multilingual-e5-large" (~2.5GB) or "intfloat/multilingual-e5-base" (~1.5GB)11# IMPORTANT: If you change the embedding model (sentence-transformers/all-MiniLM-L6-v2) and vice versa, you aren't able to use RAG Chat with your previous documents loaded in the WebUI! You need to re-embed them.12ARG USE_EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v213ARG USE_RERANKING_MODEL=""14 15# Tiktoken encoding name; models to use can be found at https://huggingface.co/models?library=tiktoken16ARG USE_TIKTOKEN_ENCODING_NAME="cl100k_base"17 18ARG BUILD_HASH=dev-build19# Override at your own risk - non-root configurations are untested20ARG UID=021ARG GID=022 23######## WebUI frontend ########24FROM --platform=$BUILDPLATFORM node:22-alpine3.20 AS build25ARG BUILD_HASH26 27WORKDIR /app28 29COPY package.json package-lock.json ./30RUN npm ci31 32COPY . .33ENV APP_BUILD_HASH=${BUILD_HASH}34RUN npm run build35 36######## WebUI backend ########37FROM python:3.11-slim-bookworm AS base38 39# Use args40ARG USE_CUDA41ARG USE_OLLAMA42ARG USE_CUDA_VER43ARG USE_EMBEDDING_MODEL44ARG USE_RERANKING_MODEL45ARG UID46ARG GID47 48## Basis ##49ENV ENV=prod \50    PORT=8080 \51    # pass build args to the build52    USE_OLLAMA_DOCKER=${USE_OLLAMA} \53    USE_CUDA_DOCKER=${USE_CUDA} \54    USE_CUDA_DOCKER_VER=${USE_CUDA_VER} \55    USE_EMBEDDING_MODEL_DOCKER=${USE_EMBEDDING_MODEL} \56    USE_RERANKING_MODEL_DOCKER=${USE_RERANKING_MODEL}57 58## Basis URL Config ##59ENV OLLAMA_BASE_URL="/ollama" \60    OPENAI_API_BASE_URL=""61 62## API Key and Security Config ##63ENV OPENAI_API_KEY="" \64    WEBUI_SECRET_KEY="" \65    SCARF_NO_ANALYTICS=true \66    DO_NOT_TRACK=true \67    ANONYMIZED_TELEMETRY=false68 69#### Other models #########################################################70## whisper TTS model settings ##71ENV WHISPER_MODEL="base" \72    WHISPER_MODEL_DIR="/app/backend/data/cache/whisper/models"73 74## RAG Embedding model settings ##75ENV RAG_EMBEDDING_MODEL="$USE_EMBEDDING_MODEL_DOCKER" \76    RAG_RERANKING_MODEL="$USE_RERANKING_MODEL_DOCKER" \77    SENTENCE_TRANSFORMERS_HOME="/app/backend/data/cache/embedding/models"78 79## Tiktoken model settings ##80ENV TIKTOKEN_ENCODING_NAME="cl100k_base" \81    TIKTOKEN_CACHE_DIR="/app/backend/data/cache/tiktoken"82 83## Hugging Face download cache ##84ENV HF_HOME="/app/backend/data/cache/embedding/models"85 86## Torch Extensions ##87# ENV TORCH_EXTENSIONS_DIR="/.cache/torch_extensions"88 89#### Other models ##########################################################90 91WORKDIR /app/backend92 93ENV HOME=/root94# Create user and group if not root95RUN if [ $UID -ne 0 ]; then \96    if [ $GID -ne 0 ]; then \97    addgroup --gid $GID app; \98    fi; \99    adduser --uid $UID --gid $GID --home $HOME --disabled-password --no-create-home app; \100    fi101 102RUN mkdir -p $HOME/.cache/chroma103RUN echo -n 00000000-0000-0000-0000-000000000000 > $HOME/.cache/chroma/telemetry_user_id104 105# Make sure the user has access to the app and root directory106RUN chown -R $UID:$GID /app $HOME107 108RUN if [ "$USE_OLLAMA" = "true" ]; then \109    apt-get update && \110    # Install pandoc and netcat111    apt-get install -y --no-install-recommends git build-essential pandoc netcat-openbsd curl && \112    apt-get install -y --no-install-recommends gcc python3-dev && \113    # for RAG OCR114    apt-get install -y --no-install-recommends ffmpeg libsm6 libxext6 && \115    # install helper tools116    apt-get install -y --no-install-recommends curl jq && \117    # install ollama118    curl -fsSL https://ollama.com/install.sh | sh && \119    # cleanup120    rm -rf /var/lib/apt/lists/*; \121    else \122    apt-get update && \123    # Install pandoc, netcat and gcc124    apt-get install -y --no-install-recommends git build-essential pandoc gcc netcat-openbsd curl jq && \125    apt-get install -y --no-install-recommends gcc python3-dev && \126    # for RAG OCR127    apt-get install -y --no-install-recommends ffmpeg libsm6 libxext6 && \128    # cleanup129    rm -rf /var/lib/apt/lists/*; \130    fi131 132# install python dependencies133COPY --chown=$UID:$GID ./backend/requirements.txt ./requirements.txt134 135RUN pip3 install uv && \136    if [ "$USE_CUDA" = "true" ]; then \137    # If you use CUDA the whisper and embedding model will be downloaded on first use138    pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/$USE_CUDA_DOCKER_VER --no-cache-dir && \139    uv pip install --system -r requirements.txt --no-cache-dir && \140    python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ['RAG_EMBEDDING_MODEL'], device='cpu')" && \141    python -c "import os; from faster_whisper import WhisperModel; WhisperModel(os.environ['WHISPER_MODEL'], device='cpu', compute_type='int8', download_root=os.environ['WHISPER_MODEL_DIR'])"; \142    python -c "import os; import tiktoken; tiktoken.get_encoding(os.environ['TIKTOKEN_ENCODING_NAME'])"; \143    else \144    pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu --no-cache-dir && \145    uv pip install --system -r requirements.txt --no-cache-dir && \146    python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ['RAG_EMBEDDING_MODEL'], device='cpu')" && \147    python -c "import os; from faster_whisper import WhisperModel; WhisperModel(os.environ['WHISPER_MODEL'], device='cpu', compute_type='int8', download_root=os.environ['WHISPER_MODEL_DIR'])"; \148    python -c "import os; import tiktoken; tiktoken.get_encoding(os.environ['TIKTOKEN_ENCODING_NAME'])"; \149    fi; \150    chown -R $UID:$GID /app/backend/data/151 152 153 154# copy embedding weight from build155# RUN mkdir -p /root/.cache/chroma/onnx_models/all-MiniLM-L6-v2156# COPY --from=build /app/onnx /root/.cache/chroma/onnx_models/all-MiniLM-L6-v2/onnx157 158# copy built frontend files159COPY --chown=$UID:$GID --from=build /app/build /app/build160COPY --chown=$UID:$GID --from=build /app/CHANGELOG.md /app/CHANGELOG.md161COPY --chown=$UID:$GID --from=build /app/package.json /app/package.json162 163# copy backend files164COPY --chown=$UID:$GID ./backend .165 166EXPOSE 8080167 168HEALTHCHECK CMD curl --silent --fail http://localhost:${PORT:-8080}/health | jq -ne 'input.status == true' || exit 1169 170USER $UID:$GID171 172ARG BUILD_HASH173ENV WEBUI_BUILD_VERSION=${BUILD_HASH}174ENV DOCKER=true175 176CMD [ "bash", "start.sh"]177