fix: install CPU-only torch to avoid 2GB+ CUDA packages on CI runner

kokoro and transformers pull full nvidia CUDA wheels by default (~2 GB),
exhausting the runner disk. Pre-installing torch from the CPU wheel index
satisfies the dependency and prevents pip from selecting the CUDA variant.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Dockerfile

@@ -16,9 +16,10 @@ COPY src/ src/
 RUN --mount=type=cache,target=/root/.cache/pip \
     pip install .
 # Voice dependencies (faster-whisper, Kokoro TTS, soundfile) — activated at runtime via VOICE_ENABLED
-# Cache mount keeps wheels across builds so the heavy torch/CUDA deps aren't re-downloaded every time.
+# Install CPU-only torch first so pip doesn't pull full CUDA wheels (~2 GB) for kokoro/transformers.
 RUN --mount=type=cache,target=/root/.cache/pip \
-    pip install faster-whisper kokoro soundfile
+    pip install torch --index-url https://download.pytorch.org/whl/cpu \
+    && pip install faster-whisper kokoro soundfile
 
 # Build the fable-mcp wheel so it can be served for download
 COPY fable-mcp/ fable-mcp/