File size: 1,584 Bytes
3be54c6 | 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 | # Multi-stage Dockerfile for tau-rag
# ---------------------------------------------------------------------------
# STAGE 1: base — slim Python + minimal deps (runs mock + extractive).
# STAGE 2: full — adds torch, faiss, sentence-transformers (AlephBERT).
#
# Usage:
# docker build -t tau-rag:slim --target slim .
# docker build -t tau-rag:full --target full .
# docker run -p 8000:8000 tau-rag:slim
# ---------------------------------------------------------------------------
FROM python:3.11-slim AS slim
WORKDIR /app
# Only what we need for the zero-dep pipeline + FastAPI
COPY pyproject.toml /app/
COPY . /app/tau_rag
RUN pip install --no-cache-dir \
"pydantic>=2.6" \
"rank-bm25>=0.2.2" \
"numpy>=1.26" \
"fastapi>=0.111" \
"uvicorn[standard]>=0.29" \
"pypdf>=3.0"
ENV PYTHONPATH=/app
EXPOSE 8000
CMD ["uvicorn", "tau_rag.api.fastapi_app:app", \
"--host", "0.0.0.0", "--port", "8000", "--workers", "2"]
# ---------------------------------------------------------------------------
FROM slim AS full
# Adds AlephBERT + FAISS — ~3GB image but real semantic retrieval.
RUN pip install --no-cache-dir \
"torch>=2.0" \
"transformers>=4.40" \
"sentence-transformers>=2.6" \
"faiss-cpu>=1.8"
# Pre-warm AlephBERT so cold-start at runtime is just a memmap open.
RUN python -c "from tau_rag.core.embedders import make_transformers_embedder; \
make_transformers_embedder('onlplab/alephbert-base')" || true
CMD ["uvicorn", "tau_rag.api.fastapi_app:app", \
"--host", "0.0.0.0", "--port", "8000", "--workers", "2"]
|