Update Dockerfile

Dockerfile

@@ -1,41 +1,63 @@
-FROM python:3.10-slim
+FROM python:3.11-slim
 
-# Install system dependencies
-RUN apt-get update && apt-get install -y \
-    libsndfile1 build-essential git curl && \
-    rm -rf /var/lib/apt/lists/*
+# HuggingFace Spaces runs as root by default on free tier
+# but expects port 7860
+ENV PORT=7860
+ENV PYTHONUNBUFFERED=1
+ENV PYTHONDONTWRITEBYTECODE=1
+ENV HF_HOME=/app/cache
+ENV TRANSFORMERS_CACHE=/app/cache
+ENV SENTENCE_TRANSFORMERS_HOME=/app/cache
 
-RUN useradd -m -u 1000 jarvis
-USER jarvis
-ENV HOME=/home/jarvis \
-    PATH=/home/jarvis/.local/bin:$PATH \
-    HF_HOME=/home/jarvis/.cache/huggingface
+WORKDIR /app
 
-WORKDIR $HOME/app
+# ── System deps needed to install pre-built llama-cpp-python wheel ──
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    curl wget ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
 
-# Force binary-only installs and use the specific wheel index for llama-cpp
-# This prevents it from even trying to compile
-ENV CMAKE_ARGS="-DLLAMA_METAL=off"
-ENV FORCE_CMAKE=1
+# ── Python deps ──
+# Install everything EXCEPT llama-cpp-python first (fast)
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
 
-RUN pip install --no-cache-dir --upgrade pip
+# ── llama-cpp-python: pre-built CPU wheel (no compile, seconds not minutes) ──
+RUN pip install --no-cache-dir llama-cpp-python \
+    --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
 
+# ── KittenTTS ──
 RUN pip install --no-cache-dir \
-    flask pydantic numpy transformers accelerate \
-    sentence-transformers faiss-cpu soundfile \
-    langchain-huggingface langchain-community \
-    langchain-text-splitters huggingface_hub
+    https://github.com/KittenML/KittenTTS/releases/download/0.8/kittentts-0.8.0-py3-none-any.whl
 
-# THIS IS THE KEY FIX: Install the specific CPU wheel directly without building
-RUN pip install --no-cache-dir llama-cpp-python \
-    --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
+# ── Pre-download the GGUF model at BUILD time so startup is instant ──
+# This bakes the model into the image layer — no download on first run
+RUN python - <<'EOF'
+from huggingface_hub import hf_hub_download
+import os
+path = hf_hub_download(
+    repo_id="unsloth/Qwen3.5-0.8B-GGUF",
+    filename="Qwen3.5-0.8B-UD-Q2_K_XL.gguf",
+    cache_dir="/app/cache"
+)
+print(f"Model cached at: {path}")
+EOF
 
-RUN pip install --no-cache-dir https://github.com/KittenML/KittenTTS/releases/download/0.8/kittentts-0.8.0-py3-none-any.whl
+# ── Pre-download embeddings model at BUILD time ──
+RUN python - <<'EOF'
+from sentence_transformers import SentenceTransformer
+SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2",
+                    cache_folder="/app/cache")
+print("Embeddings model cached.")
+EOF
 
-COPY --chown=jarvis:jarvis . $HOME/app
+# ── Copy app source ──
+COPY app.py .
+COPY static/ static/
 
-# Pre-download models
-RUN python app.py --setup
+# ── Create runtime dirs ──
+RUN mkdir -p /app/database/vector_store \
+             /app/database/learning_data \
+             /app/database/chats_data
 
 EXPOSE 7860