feat: Dockerfile, model download script, and pod creation script

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

kischdle/llmux/scripts/download_models.sh

@@ -0,0 +1,72 @@
+#!/bin/bash
+# Download all model weights for llmux.
+# Run as user llm: bash scripts/download_models.sh
+# Requires: pip install huggingface_hub
+# Requires: HuggingFace token at ~/.cache/huggingface/token for gated models
+
+set -euo pipefail
+
+MODELS_DIR="${LLMUX_MODELS_DIR:-$HOME/.local/share/llmux_pod/models}"
+mkdir -p "$MODELS_DIR"
+
+echo "=== Downloading models to $MODELS_DIR ==="
+
+download_hf() {
+    local repo="$1"
+    local target="$MODELS_DIR/models--${repo//\//-}"
+    if [ -d "$target" ]; then
+        echo "SKIP: $repo (already downloaded)"
+        return
+    fi
+    echo "Downloading: $repo"
+    huggingface-cli download "$repo" --cache-dir "$MODELS_DIR"
+}
+
+download_hf_files() {
+    local repo="$1"
+    shift
+    echo "Downloading specific files from: $repo"
+    huggingface-cli download "$repo" "$@" --cache-dir "$MODELS_DIR"
+}
+
+# 1. Qwen3.5-9B-FP8
+download_hf "lovedheart/Qwen3.5-9B-FP8"
+
+# 2. Qwen3.5-9B-FP8-Uncensored (GGUF files only)
+download_hf_files "HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive" \
+    "Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-Q8_0.gguf" \
+    "mmproj-Qwen3.5-9B-Uncensored-HauhauCS-Aggressive-BF16.gguf"
+
+# 3. Qwen3.5-4B
+download_hf "Qwen/Qwen3.5-4B"
+
+# 4. gpt-oss-20b
+download_hf "openai/gpt-oss-20b"
+
+# 5. gpt-oss-20b-uncensored
+download_hf "aoxo/gpt-oss-20b-uncensored"
+
+# 6. cohere-transcribe (gated — requires accepted terms)
+echo "Downloading: CohereLabs/cohere-transcribe-03-2026 (gated)"
+download_hf "CohereLabs/cohere-transcribe-03-2026" || \
+    echo "WARNING: cohere-transcribe download failed. Have you accepted the terms at https://huggingface.co/CohereLabs/cohere-transcribe-03-2026 ?"
+
+# 7. Chatterbox TTS
+echo "Downloading: Chatterbox TTS weights (auto-downloaded by library)"
+python3 -c "
+from chatterbox.tts import ChatterboxTTS
+import os
+os.environ['CUDA_VISIBLE_DEVICES'] = ''
+print('Downloading Chatterbox default...')
+ChatterboxTTS.from_pretrained(device='cpu')
+print('Downloading Chatterbox turbo...')
+ChatterboxTTS.from_pretrained(device='cpu', variant='turbo')
+print('Downloading Chatterbox multilingual...')
+ChatterboxTTS.from_pretrained(device='cpu', variant='multilingual')
+print('Chatterbox downloads complete.')
+" || echo "WARNING: Chatterbox download failed. Check chatterbox-tts installation."
+
+echo ""
+echo "=== Download complete ==="
+echo "Models directory: $MODELS_DIR"
+du -sh "$MODELS_DIR"