digraph AutoAuthorRecognition {
    rankdir=LR;
    node [shape=box, style=filled, fillcolor=lightgray];

    subgraph cluster_api {
        label = "FastAPI Backend";
        style=filled;
        color=lightblue;
        main_py [label="main.py\n(FastAPI app)"];
        lstm_inference [label="lstm_inference.py\n(LSTM logic)"];
        transformer_inference [label="transformer_inference.py\n(Transformer logic)"];
        main_py -> lstm_inference [label="/predict/lstm\nimport & call"];
        main_py -> transformer_inference [label="/predict/transformer\nimport & call"];
        transformer_inference;
    }

    subgraph cluster_lstm {
        label = "LSTM Module";
        style=filled;
        color=lightyellow;
        lstm_inference;
        lstm_weights [label="weights/\nLSTM model weights", shape=note, fillcolor=white];
        lstm_inference -> lstm_weights [label="load weights"];
    }

    subgraph cluster_transformer {
        label = "Transformer Module";
        style=filled;
        color=lightyellow;
        transformer_weights [label="weights/\nTransformer model files", shape=note, fillcolor=white];
        transformer_inference -> transformer_weights [label="load weights"];
    }

    subgraph cluster_frontend {
        label = "Streamlit Frontend";
        style=filled;
        color=lightgreen;
        streamlit_app [label="streamlit_app.py\n(UI)"];
        streamlit_app -> main_py [label="API calls"];
    }

    user [label="User", shape=ellipse, fillcolor=white];
    user -> streamlit_app [label="Web UI"];
    user -> main_py [label="API (optional)"];
}