Spaces:
Sleeping
Sleeping
File size: 13,422 Bytes
ce11d27 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 | """
3D scatter plot with Catmull-Rom spline paths β faithful port from Android.
Features ported from My3DScatterRenderer.java:
- Catmull-Rom spline interpolation (20 samples/segment, exact b0-b3 coefficients)
- Android cluster color palette (10 colors)
- Dark background (#1E1E1E matching Android's glClearColor 0.12, 0.12, 0.12)
- Proper point sizing
"""
from __future__ import annotations
from typing import List, Optional
import numpy as np
# Plotly is only needed for the plotting functions, not for catmull_rom_spline.
# Defer import so gl_renderer can import catmull_rom_spline without plotly.
go = None # lazy-loaded
def _require_plotly():
global go
if go is not None:
return
try:
import plotly.graph_objects as _go
go = _go
except ImportError as _e:
raise ImportError(
"Visualization dependencies not installed. "
"Install them with: pip install plotly\n"
"Or install the full package: pip install tracescope"
) from _e
from tracescope.models.analysis import AnalysisResult
# βββββββββββββββββββββββββββββββββββββββββββββββββββ
# Android cluster palette (exact RGB from DashboardFragment.java line 2254)
# βββββββββββββββββββββββββββββββββββββββββββββββββββ
CLUSTER_COLORS = [
"rgb(255, 0, 0)", # red
"rgb(0, 255, 0)", # green
"rgb(0, 0, 255)", # blue
"rgb(255, 255, 0)", # yellow
"rgb(255, 0, 255)", # magenta
"rgb(0, 255, 255)", # cyan
"rgb(255, 128, 0)", # orange
"rgb(128, 0, 255)", # purple
"rgb(0, 128, 255)", # sky blue
"rgb(128, 255, 128)", # pastel green
]
# Hex versions for use where needed
CLUSTER_COLORS_HEX = [
"#FF0000", "#00FF00", "#0000FF", "#FFFF00", "#FF00FF",
"#00FFFF", "#FF8000", "#8000FF", "#0080FF", "#80FF80",
]
def catmull_rom_spline(
points: np.ndarray,
samples_per_segment: int = 20,
) -> np.ndarray:
"""Catmull-Rom spline interpolation.
Exact coefficients from Android My3DScatterRenderer.java lines 414-456:
b0 = -0.5*tΒ³ + tΒ² - 0.5*t
b1 = 1.5*tΒ³ - 2.5*tΒ² + 1
b2 = -1.5*tΒ³ + 2.0*tΒ² + 0.5*t
b3 = 0.5*tΒ³ - 0.5*tΒ²
Mirrors first/last control points for boundary handling.
Args:
points: (N, 3) array of control points.
samples_per_segment: Number of interpolated samples per segment (default 20).
Returns:
(M, 3) array of interpolated points.
"""
N = len(points)
if N < 2:
return points.copy()
if N == 2:
# Simple linear interpolation
result = []
for j in range(samples_per_segment + 1):
t = j / samples_per_segment
result.append(points[0] * (1 - t) + points[1] * t)
return np.array(result)
# Mirror endpoints for boundary handling
# p_extended = [mirror_start, p0, p1, ..., pN-1, mirror_end]
mirror_start = 2 * points[0] - points[1]
mirror_end = 2 * points[-1] - points[-2]
extended = np.vstack([mirror_start, points, mirror_end])
result = []
# Iterate over segments: for each group of 4 consecutive points (p0, p1, p2, p3)
# the spline passes through p1 β p2
for seg in range(len(extended) - 3):
p0 = extended[seg]
p1 = extended[seg + 1]
p2 = extended[seg + 2]
p3 = extended[seg + 3]
for j in range(samples_per_segment):
t = j / samples_per_segment
t2 = t * t
t3 = t2 * t
# Catmull-Rom basis functions (exact Android coefficients)
b0 = -0.5 * t3 + t2 - 0.5 * t
b1 = 1.5 * t3 - 2.5 * t2 + 1.0
b2 = -1.5 * t3 + 2.0 * t2 + 0.5 * t
b3 = 0.5 * t3 - 0.5 * t2
point = b0 * p0 + b1 * p1 + b2 * p2 + b3 * p3
result.append(point)
# Add the last point
result.append(extended[-2]) # Last original point
return np.array(result)
def _apply_dark_theme(fig, axis_labels: List[str]) -> None:
"""Apply dark theme matching Android's glClearColor(0.12, 0.12, 0.12)."""
bg_color = "rgb(30, 30, 30)"
grid_color = "rgb(60, 60, 60)"
axis_common = dict(
backgroundcolor=bg_color,
gridcolor=grid_color,
color="white",
showbackground=True,
zerolinecolor=grid_color,
)
fig.update_layout(
scene=dict(
bgcolor=bg_color,
xaxis=dict(title=axis_labels[0] if len(axis_labels) > 0 else "X", **axis_common),
yaxis=dict(title=axis_labels[1] if len(axis_labels) > 1 else "Y", **axis_common),
zaxis=dict(title=axis_labels[2] if len(axis_labels) > 2 else "Z", **axis_common),
),
paper_bgcolor=bg_color,
plot_bgcolor=bg_color,
font=dict(color="white"),
legend=dict(font=dict(color="white")),
)
def _score_to_plotly_colors(values: List[Optional[float]], fallback_color: str = "rgb(128,128,128)") -> List[str]:
"""Convert score values to plotly color strings using red-yellow-green gradient."""
from tracescope.visualization.flow_field import score_colormap
colors = []
for v in values:
if v is None:
colors.append(fallback_color)
else:
rgb = score_colormap(np.array([v]))[0]
colors.append(f"rgb({int(rgb[0]*255)},{int(rgb[1]*255)},{int(rgb[2]*255)})")
return colors
def plot_clusters_3d(
result: AnalysisResult,
show_path: bool = True,
marker_size: int = 6,
path_width: int = 5,
use_spline: bool = True,
color_by_score: Optional[str] = None,
):
"""Create a 3D scatter plot with clusters colored and Catmull-Rom spline path.
Args:
result: AnalysisResult from the pipeline.
show_path: Whether to draw the conversation path.
marker_size: Size of scatter points.
path_width: Width of the path line.
use_spline: Use Catmull-Rom spline (True) or straight lines (False).
color_by_score: Optional score channel name to color points by score
instead of cluster. Uses red-yellow-green gradient.
Returns:
Plotly Figure object.
"""
_require_plotly()
fig = go.Figure()
pts = result.projected_3d
labels = result.clusters.labels
axis_labels = result.axis_info.labels
n_clusters = result.clusters.n_clusters
if color_by_score and color_by_score in result.score_channels:
# ββ Score-based coloring ββββββββββββββββββββββββββ
entry_scores = result.get_entry_scores(color_by_score)
path_score_map = result.get_path_scores(color_by_score)
# For each entry: use entry score if present, else path score, else None
final_scores = []
for i, e in enumerate(result.session.entries):
s = entry_scores[i]
if s is None and e.path_id is not None:
s = path_score_map.get(e.path_id)
final_scores.append(s)
point_colors = _score_to_plotly_colors(final_scores)
hover = [
f"[{result.session.entries[i].role}] {result.session.entries[i].text[:100]}... "
f"({color_by_score}: {final_scores[i]:.2f})" if final_scores[i] is not None
else f"[{result.session.entries[i].role}] {result.session.entries[i].text[:100]}..."
for i in range(len(pts))
]
fig.add_trace(
go.Scatter3d(
x=pts[:, 0], y=pts[:, 1], z=pts[:, 2],
mode="markers",
marker=dict(size=marker_size, color=point_colors, opacity=0.8),
name=f"Score: {color_by_score}",
text=hover,
hoverinfo="text",
)
)
else:
# ββ Cluster-based coloring (default) ββββββββββββββ
for c in range(n_clusters):
mask = [i for i, l in enumerate(labels) if l == c]
if not mask:
continue
cluster_pts = pts[mask]
cluster_texts = [result.session.entries[i].text[:100] for i in mask]
cluster_roles = [result.session.entries[i].role for i in mask]
hover = [
f"[{role}] {text}..."
for role, text in zip(cluster_roles, cluster_texts)
]
color = CLUSTER_COLORS[c % len(CLUSTER_COLORS)]
label = (
result.cluster_labels[c]
if c < len(result.cluster_labels)
else f"Cluster {c}"
)
fig.add_trace(
go.Scatter3d(
x=cluster_pts[:, 0],
y=cluster_pts[:, 1],
z=cluster_pts[:, 2],
mode="markers",
marker=dict(size=marker_size, color=color, opacity=0.8),
name=label,
text=hover,
hoverinfo="text",
)
)
# Path line (Catmull-Rom spline or straight)
if show_path and len(pts) >= 2:
if use_spline:
spline_pts = catmull_rom_spline(pts, samples_per_segment=20)
else:
spline_pts = pts
# Color path by score if available
if color_by_score and color_by_score in result.score_channels:
path_score_map = result.get_path_scores(color_by_score)
entry_scores = result.get_entry_scores(color_by_score)
# Build per-entry score, falling back to path score
per_entry = []
for i, e in enumerate(result.session.entries):
s = entry_scores[i]
if s is None and e.path_id is not None:
s = path_score_map.get(e.path_id)
per_entry.append(s if s is not None else 0.5)
path_colors = _score_to_plotly_colors(per_entry)
# For spline: repeat colors for interpolated points
if use_spline and len(per_entry) >= 2:
n_seg = len(per_entry) - 1
expanded = []
for seg_i in range(n_seg):
for j in range(20):
t = j / 20.0
v = per_entry[seg_i] * (1 - t) + per_entry[seg_i + 1] * t
expanded.append(v)
expanded.append(per_entry[-1])
# Trim to match spline length
expanded = expanded[:len(spline_pts)]
while len(expanded) < len(spline_pts):
expanded.append(per_entry[-1])
path_colors = _score_to_plotly_colors(expanded)
fig.add_trace(
go.Scatter3d(
x=spline_pts[:, 0], y=spline_pts[:, 1], z=spline_pts[:, 2],
mode="lines",
line=dict(color=path_colors, width=path_width),
name="Path",
showlegend=True,
hoverinfo="skip",
)
)
else:
fig.add_trace(
go.Scatter3d(
x=spline_pts[:, 0],
y=spline_pts[:, 1],
z=spline_pts[:, 2],
mode="lines",
line=dict(color="rgba(255,255,255,0.7)", width=path_width),
name="Path",
showlegend=True,
hoverinfo="skip",
)
)
# Apply dark theme
_apply_dark_theme(fig, axis_labels)
fig.update_layout(
title="TraceScope: Conversation in Semantic Space",
showlegend=True,
margin=dict(l=0, r=0, t=40, b=0),
)
return fig
def plot_multi_paths(
results: List[AnalysisResult],
labels: Optional[List[str]] = None,
):
"""Overlay multiple conversation paths in the same 3D space."""
_require_plotly()
fig = go.Figure()
for i, result in enumerate(results):
pts = result.projected_3d
label = labels[i] if labels and i < len(labels) else result.session.label
color = CLUSTER_COLORS[i % len(CLUSTER_COLORS)]
# Spline path
if len(pts) >= 2:
spline_pts = catmull_rom_spline(pts, samples_per_segment=20)
else:
spline_pts = pts
fig.add_trace(
go.Scatter3d(
x=spline_pts[:, 0],
y=spline_pts[:, 1],
z=spline_pts[:, 2],
mode="lines",
line=dict(color=color, width=4),
name=label,
)
)
# Add scatter points
fig.add_trace(
go.Scatter3d(
x=pts[:, 0],
y=pts[:, 1],
z=pts[:, 2],
mode="markers",
marker=dict(size=4, color=color, opacity=0.7),
name=f"{label} (points)",
showlegend=False,
)
)
axis_labels = results[0].axis_info.labels if results else ["X", "Y", "Z"]
_apply_dark_theme(fig, axis_labels)
fig.update_layout(
title="TraceScope: Multiple Conversation Paths",
showlegend=True,
)
return fig
|