w

Dockerfile

@@ -2,36 +2,29 @@
 # This is a pre-built image with everything included.
 FROM ghcr.io/huggingface/text-generation-inference:latest
 
-# --- 2. Install Python & venv to download our model ---
-# We still need to download our model, so we add Python.
-RUN apt-get update && \
-    apt-get install -y python3 python3-pip python3-venv && \
-    rm -rf /var/lib/apt/lists/*
-
-# --- 3. Download the GGUF model ---
+# --- 2. Download the GGUF model using cURL ---
+# We use cURL (which is already in the image) to avoid
+# installing Python and causing version conflicts.
 WORKDIR /app
-COPY download_model.py .
 
+# Get the HF_TOKEN from the build secrets
 ARG HF_TOKEN
-# This command creates a venv, installs hf_hub, downloads the model,
-# and then the venv is discarded.
+
+# Run the download command
 RUN --mount=type=secret,id=HF_TOKEN \
-    sh -c 'python3 -m venv /tmp/downloader-venv && \
-           . /tmp/downloader-venv/bin/activate && \
-           pip install huggingface_hub && \
-           python3 download_model.py'
+    curl -L \
+         -H "Authorization: Bearer $(cat /run/secrets/HF_TOKEN)" \
+         "https://huggingface.co/mradermacher/prem-1B-SQL-GGUF/resolve/main/prem-1B-SQL.Q8_0.gguf" \
+         -o "prem-1B-SQL.Q8_0.gguf"
 
-# --- 4. Set the container's command to run TGI ---
+# --- 3. Set the container's command to run TGI ---
 # This is the command that will run when the container starts.
-# It tells TGI to serve our GGUF model and to protect
-# the API with the key we set in our secrets.
 ENV MODEL_ID="/app/prem-1B-SQL.Q8_0.gguf"
 
 CMD [ \
     "text-generation-launcher", \
     "--model-id", "${MODEL_ID}", \
     "--quantize", "gguf", \
-    # This is the fix: Changed from 80 to 8000
     "--port", "8000", \
     "--host", "0.0.0.0", \
     "--openai-api-key-env-var", "API_KEY" \