bit_transformer/dashboard.py

import matplotlib.pyplot as plt
from typing import Dict, List, Tuple


def plot_telemetry(
    metrics_log: Dict[str, List[float]],
    k_floor: float = 0.5,
    c_floor: float = 0.3,
    s_floor: float = 0.5,
) -> Tuple[plt.Figure, List[plt.Axes]]:
    """Plot K, C, S metrics over time with cluster transitions.

    Args:
        metrics_log: Dictionary with keys ``negentropy``, ``lz_complexity``,
            ``symbiosis_score`` and optional ``clusters`` listing cluster
            assignments per step.
        k_floor: Threshold for negentropy (K).
        c_floor: Threshold for LZ complexity (C).
        s_floor: Threshold for symbiosis score (S).

    Returns:
        (figure, axes) tuple for further customization or saving.
    """
    steps = list(range(len(metrics_log.get("negentropy", []))))
    fig, axes = plt.subplots(3, 1, sharex=True, figsize=(10, 6))
    metrics = [
        ("negentropy", k_floor, "K"),
        ("lz_complexity", c_floor, "C"),
        ("symbiosis_score", s_floor, "S"),
    ]
    for ax, (key, floor, label) in zip(axes, metrics):
        values = metrics_log.get(key, [])
        ax.plot(steps, values, label=label)
        ax.axhline(floor, color="r", linestyle="--", linewidth=1)
        violations = [i for i, v in enumerate(values) if v < floor]
        if violations:
            ax.scatter(
                [steps[i] for i in violations],
                [values[i] for i in violations],
                color="r",
                zorder=5,
                label="violation",
            )
        ax.set_ylabel(label)
        ax.legend(loc="upper right")

    clusters = metrics_log.get("clusters")
    if clusters is not None:
        prev = clusters[0]
        for t, c in enumerate(clusters):
            if t > 0 and c != prev:
                for ax in axes:
                    ax.axvline(t, color="gray", linestyle=":", alpha=0.5)
                prev = c

    axes[-1].set_xlabel("step")
    plt.tight_layout()
    return fig, axes