How do I convert RGB to Oklab?

[test3211234]

Question

I have a list of RGB tuples. I want to make it Oklab and make a linear gradient with it, then turn it into an RGB Pillow Image. Regular code for RGB looks like this. Please critique also, it's supposed to be as fast as it can be.

def create_linear_gradient(width, height, stops, direction):
    # stops is a list { percent, (r, g, b) }

    # Normalize the direction vector
    dx, dy = direction
    magnitude = (dx ** 2 + dy ** 2) ** 0.5
    dx /= magnitude
    dy /= magnitude

    # create coordinate grids
    x = np.linspace(0, 1, width)
    y = np.linspace(0, 1, height)
    xv, yv = np.meshgrid(x, y)

    # Compute the distance map and center it
    distance_map = (xv - 0.5) * dx + (yv - 0.5) * dy
    distance_map = distance_map - distance_map.min()
    distance_map = distance_map / distance_map.max()  # Normalize to [0, 1]

    # Sort and interpolate stops
    stops = sorted(stops, key=lambda stop: stop[1])
    stop_colors, stop_positions = zip(*stops)

    # Convert to arrays for interpolation
    stop_positions = np.array(stop_positions)
    stop_colors = np.array(stop_colors, dtype=np.float32)

    # Interpolate each color channel
    r_channel = np.interp(distance_map, stop_positions, stop_colors[:, 0])
    g_channel = np.interp(distance_map, stop_positions, stop_colors[:, 1])
    b_channel = np.interp(distance_map, stop_positions, stop_colors[:, 2])

    gradient = np.stack([r_channel, g_channel, b_channel], axis=-1).astype(np.uint8)

    return Image.fromarray(gradient)

[KelSolaar]

Typically, if you want a perceptually linear gradient, you would construct it within Oklab, idea would be to have your input values as RGB, e.g., start and end RGB triplets, convert them to Oklab and linear interpolate the Oklab correlates.