config.yaml

macros:
  "latest-llama": >
    /app/llama.cpp/build/bin/llama-server
    --port ${PORT}
    --no-mmap
    --threads 2

models:
  "Ernie-4.5-0.3B":
    cmd: |
      ${latest-llama}
      -m /app/llama.cpp/build/ERNIE-4.5-0.3B-PT-UD-Q8_K_XL.gguf
      -c 32000
      --jinja
    ttl: 360
  "LFM2-VL-450M":
    cmd: |
      ${latest-llama}
      -m /app/llama.cpp/build/LFM2-VL-450M-Q8_0.gguf
      --mmproj /app/llama.cpp/build/mmproj-LFM2-VL-450M-F16.gguf
      -c 32000
    ttl: 360
  "gemma-3-270m-it":
    cmd: |
      ${latest-llama}
      -m /app/llama.cpp/build/gemma-3-270m-it-UD-Q8_K_XL.gguf
      -c 32000
      --jinja
    ttl: 360
  "Qwen3-0.6B":
    cmd: |
      ${latest-llama}
      -m /app/llama.cpp/build/Qwen3-0.6B-UD-Q8_K_XL.gguf
      -c 32000
      --jinja
    ttl: 360