@pentatonic-ai/ai-agent-sdk 0.9.6 → 0.10.0

package/packages/memory-engine/engine/services/l6/Dockerfile

@@ -1,30 +0,0 @@
-FROM python:3.12-slim
-WORKDIR /app
-# gcc/g++ needed by some sentence-transformers transitive deps; curl kept
-# for in-container debugging.
-RUN apt-get update && apt-get install -y --no-install-recommends \
-        gcc g++ curl \
-    && rm -rf /var/lib/apt/lists/*
-# Core deps + sentence-transformers/torch for the cross-encoder reranker.
-# Torch CPU wheel is enough — the reranker is small (MiniLM L-6) and
-# CPU-bound throughput is fine at L6's request volume. Without these,
-# get_reranker() falls back to RRF-only, capping recall ranking quality.
-RUN pip install --no-cache-dir \
-        fastapi "uvicorn[standard]" httpx "pymilvus[milvus_lite]" "setuptools<70" pydantic spacy \
-        "sentence-transformers" \
-        "torch" --extra-index-url https://download.pytorch.org/whl/cpu
-RUN python -m spacy download en_core_web_sm
-ENV HF_HOME=/data/.cache/huggingface
-# Pre-download the cross-encoder so cold-start doesn't pay first-pull
-# latency. The model is small (~80MB) and gets cached at /data — survives
-# container recreates since /data is a volume mount.
-RUN mkdir -p /data/.cache/huggingface && \
-    python -c "from sentence_transformers import CrossEncoder; CrossEncoder('cross-encoder/ms-marco-MiniLM-L-6-v2', cache_folder='/data/.cache/huggingface')"
-# Shared embed_provider module (build context is engine/services).
-COPY _shared /app/_shared
-COPY l6/l6-document-store.py /app/server.py
-RUN mkdir -p /data
-ENV L6_DATA_DIR=/data
-EXPOSE 8037
-ENV HOST=0.0.0.0
-CMD ["python", "server.py", "serve", "--port", "8037"]