mirror of
https://github.com/tiennm99/DocsGPT.git
synced 2026-10-05 10:13:39 +00:00
Alex
de22be5a21
Follow-up review pass over the embeddings branch. - Fold an oversized header back into the body, and drop header duplication when it would leave under a quarter of the chunk budget. A header at or over max_tokens collapsed the body budget to one token, so a document became one chunk per body token, each still over the cap: a 95 KB file produced 20k chunks of 2563 tokens against a 1250 cap. Also clamp max_tokens to at least 1, as the strategy chunkers already do. - Emit a header-only document as its own chunk. With no body piece to attach it to, splitting returned nothing and the document was dropped from the index with no error and no log line. - Skip add_custom_model for a repository FastEmbed already ships. It rejects a name it knows, so configuring any of its ~30 built-ins (MiniLM, bge, e5, gte, ...) failed every embed call and every query. - Decide "the user chose this model" by comparing against the field default rather than model_fields_set, which is true for anything read from .env. Every setup script has always written EMBEDDINGS_NAME, so an upgraded remote-embeddings install inherited mpnet's 384-token window and silently clipped ~80% off every chunk. - Cut tiktoken splits at character offsets instead of decoding each token window. A multi-byte character straddling a boundary decoded to U+FFFD on both sides, destroying one character at roughly one boundary in five on CJK text -- including at the default max_tokens of 2000. - Let the re-embed script open a FAISS index whose width does not match the configured model. That mismatch is the main reason to run it, and the error recommending the script was raised by the script itself, so the advice failed on every source. - Re-embed graph_nodes.name_embedding when GraphRAG is enabled. Those vectors seed every traversal and share the chunk vectors' width, so a same-width model swap left the graph retrieving from the old space with nothing to report it. - Prefetch the models before copying the application source, so editing any file no longer re-downloads ~780 MB of artifacts on every build. - Mirror the setup.sh embedding menu into setup.ps1: granite default, legacy mpnet as an explicit option, and both engine flows updated. Windows users were otherwise stranded on mpnet with no granite path. - Drop the unused EmbeddingsWrapper.tokenizer property.
126 lines
4.2 KiB
Docker
126 lines
4.2 KiB
Docker
# Builder Stage
|
|
FROM ubuntu:24.04 as builder
|
|
|
|
ENV DEBIAN_FRONTEND=noninteractive
|
|
|
|
RUN apt-get update && \
|
|
apt-get install -y software-properties-common && \
|
|
add-apt-repository ppa:deadsnakes/ppa && \
|
|
apt-get update && \
|
|
apt-get install -y --no-install-recommends gcc g++ wget unzip libc6-dev python3.12 python3.12-venv python3.12-dev && \
|
|
rm -rf /var/lib/apt/lists/*
|
|
|
|
# Verify Python installation and setup symlink
|
|
RUN if [ -f /usr/bin/python3.12 ]; then \
|
|
ln -s /usr/bin/python3.12 /usr/bin/python; \
|
|
else \
|
|
echo "Python 3.12 not found"; exit 1; \
|
|
fi
|
|
|
|
# Install Rust
|
|
RUN wget -q -O - https://sh.rustup.rs | sh -s -- -y
|
|
|
|
# Clean up to reduce container size
|
|
RUN apt-get remove --purge -y wget unzip && apt-get autoremove -y && rm -rf /var/lib/apt/lists/*
|
|
|
|
# Copy requirements.txt
|
|
COPY requirements.txt .
|
|
|
|
# Setup Python virtual environment
|
|
RUN python3.12 -m venv /venv
|
|
|
|
# Activate virtual environment and install Python packages
|
|
ENV PATH="/venv/bin:$PATH"
|
|
|
|
# Install Python packages
|
|
RUN pip install --no-cache-dir --upgrade pip && \
|
|
pip install --no-cache-dir tiktoken && \
|
|
pip install --no-cache-dir -r requirements.txt
|
|
|
|
|
|
# Final Stage
|
|
FROM ubuntu:24.04 as final
|
|
|
|
RUN apt-get update && \
|
|
apt-get install -y software-properties-common && \
|
|
add-apt-repository ppa:deadsnakes/ppa && \
|
|
apt-get update && apt-get install -y --no-install-recommends \
|
|
python3.12 \
|
|
libgl1 \
|
|
libglib2.0-0 \
|
|
poppler-utils \
|
|
&& \
|
|
ln -s /usr/bin/python3.12 /usr/bin/python && \
|
|
rm -rf /var/lib/apt/lists/*
|
|
|
|
# Set working directory
|
|
WORKDIR /app
|
|
|
|
# Create a non-root user: `appuser` (Feel free to choose a name)
|
|
RUN groupadd -r appuser && \
|
|
useradd -r -g appuser -d /app -s /sbin/nologin -c "Docker image user" appuser
|
|
|
|
# Copy the virtual environment and model from the builder stage
|
|
COPY --from=builder /venv /venv
|
|
|
|
# Pre-fetch the embedding models into FastEmbed's cache so a fresh container
|
|
# does not download on first ingest and an air-gapped install works at all.
|
|
# Both defaults are baked: an upgraded deployment keeps using mpnet until it
|
|
# runs the re-embed script, while a new one starts on granite.
|
|
# The prefetch writes hub-layout snapshots (including tokenizer.json) here, so
|
|
# HF_HUB_CACHE has to point at the same directory: chunking loads the tokenizer
|
|
# through ``tokenizers``, which reads the hub cache and would otherwise fetch
|
|
# over the network on first ingest -- and fall back to cl100k when offline.
|
|
ENV EMBEDDINGS_CACHE_DIR=/app/models \
|
|
HF_HUB_CACHE=/app/models
|
|
|
|
# Only the modules the prefetch imports are copied first. It reaches nothing
|
|
# beyond model_registry, which is stdlib-only, so keeping the full source copy
|
|
# below this layer stops an unrelated edit from re-downloading ~780 MB of model
|
|
# artifacts on every build.
|
|
COPY __init__.py /app/application/__init__.py
|
|
COPY scripts/__init__.py scripts/prefetch_models.py /app/application/scripts/
|
|
COPY vectorstore/__init__.py vectorstore/model_registry.py /app/application/vectorstore/
|
|
ARG EMBEDDINGS_PREFETCH=""
|
|
RUN PYTHONPATH=/app /venv/bin/python -m application.scripts.prefetch_models ${EMBEDDINGS_PREFETCH}
|
|
|
|
# Copy your application code
|
|
COPY . /app/application
|
|
|
|
# Change the ownership of the /app directory to the appuser
|
|
|
|
RUN mkdir -p /app/application/inputs/local
|
|
RUN chown -R appuser:appuser /app
|
|
|
|
# Set environment variables
|
|
ENV FLASK_APP=app.py \
|
|
FLASK_DEBUG=true \
|
|
PATH="/venv/bin:$PATH"
|
|
|
|
ENV MALLOC_ARENA_MAX=2 \
|
|
OMP_NUM_THREADS=4 \
|
|
MKL_NUM_THREADS=4 \
|
|
OPENBLAS_NUM_THREADS=4
|
|
|
|
# Expose the port the app runs on
|
|
EXPOSE 7091
|
|
|
|
# Switch to non-root user
|
|
USER appuser
|
|
|
|
# BoundedDrainUvicornWorker makes max_requests recycles safe with held-open SSE
|
|
# connections (see application/gunicorn_worker.py); with recycles now safe,
|
|
# --max-requests is raised (kept for memory hygiene) to cut churn.
|
|
CMD ["gunicorn", \
|
|
"-w", "1", \
|
|
"-k", "application.gunicorn_worker.BoundedDrainUvicornWorker", \
|
|
"--bind", "0.0.0.0:7091", \
|
|
"--timeout", "180", \
|
|
"--graceful-timeout", "120", \
|
|
"--keep-alive", "5", \
|
|
"--worker-tmp-dir", "/dev/shm", \
|
|
"--max-requests", "5000", \
|
|
"--max-requests-jitter", "500", \
|
|
"--config", "application/gunicorn_conf.py", \
|
|
"application.asgi:asgi_app"]
|