models_config.json

{
  "patient_summary_models": [
    {
      "name": "facebook/bart-large-cnn",
      "type": "summarization",
      "is_active": false,
      "cached": true,
      "deprecated": true,
      "description": "BART Large CNN for summarization",
      "use_case": "General text summarization",
      "notes": "⚠️ NOT RECOMMENDED FOR MEDICAL TEXT. This model was trained on news articles (CNN/DailyMail), not medical data. May produce suboptimal results for clinical text. Use Phi-3-mini-4k-instruct-q4.gguf for better medical text summarization.",
      "warning": "Limited medical domain knowledge - trained on news articles"
    },
    {
      "name": "patrickvonplaten/longformer2roberta-cnn_dailymail-fp16",
      "type": "seq2seq",
      "is_active": false,
      "cached": true,
      "deprecated": true,
      "description": "Longformer2Roberta for long document summarization",
      "use_case": "Long document processing",
      "notes": "⚠️ DEPRECATED - NOT RECOMMENDED FOR MEDICAL TEXT. This model was trained on news articles (CNN/DailyMail), not medical data, and produces irrelevant summaries for clinical text. The model fundamentally mismatches medical terminology, structured visit data, and clinical narrative style. Use Phi-3-mini-4k-instruct-q4.gguf instead.",
      "warning": "DO NOT USE - Trained on news articles, produces irrelevant medical summaries"
    },
    {
      "name": "microsoft/Phi-3-mini-4k-instruct",
      "type": "causal-openvino",
      "is_active": false,
      "cached": true,
      "description": "Phi-3 Mini base model for OpenVINO",
      "use_case": "Patient summary generation with OpenVINO optimization"
    },
    {
      "name": "OpenVINO/Phi-3-mini-4k-instruct-fp16-ov",
      "type": "causal-openvino",
      "is_active": false,
      "cached": true,
      "description": "Phi-3 Mini FP16 optimized for OpenVINO",
      "use_case": "Patient summary generation with FP16 optimization"
    },
    {
      "name": "microsoft/Phi-3-mini-4k-instruct-gguf/Phi-3-mini-4k-instruct-q4.gguf",
      "type": "gguf",
      "is_active": true,
      "cached": true,
      "description": "Phi-3 Mini GGUF Q4 quantized - PRIMARY MODEL",
      "use_case": "Fast patient summary generation with CPU/GPU",
      "repo_id": "microsoft/Phi-3-mini-4k-instruct-gguf",
      "filename": "Phi-3-mini-4k-instruct-q4.gguf"
    },
    {
      "name": "google/flan-t5-large",
      "type": "summarization",
      "is_active": false,
      "cached": true,
      "description": "FLAN-T5 Large for summarization",
      "use_case": "Alternative summarization model"
    }
  ],
  "runtime_behavior": {
    "allow_runtime_downloads": true,
    "cache_runtime_downloads": true,
    "fallback_to_cached": true,
    "description": "System will download any requested model at runtime if not cached"
  },
  "cache_directories": {
    "HF_HOME": "/app/.cache/huggingface",
    "MODEL_CACHE_DIR": "/app/models",
    "TORCH_HOME": "/app/.cache/torch",
    "WHISPER_CACHE": "/app/.cache/whisper"
  },
  "notes": [
    "Models with 'cached: true' are pre-downloaded during Docker build",
    "Models with 'is_active: true' are the primary/default models",
    "Other models can be requested at runtime and will be downloaded automatically",
    "Runtime downloads are cached for subsequent uses"
  ]
}