Spaces:
Sleeping
Sleeping
File size: 3,253 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 | """
Interactive probe – explore 3D semantic space with sliders.
"""
from __future__ import annotations
from typing import List
import numpy as np
from tracescope.models.analysis import AnalysisResult
from tracescope.analysis.explainer import SemanticExplainer
def probe_point(
result: AnalysisResult,
x: float,
y: float,
z: float,
k_nearest: int = 5,
) -> dict:
"""Given a 3D coordinate, return nearest texts, cluster info, and axis values.
Args:
result: AnalysisResult from the pipeline.
x, y, z: 3D coordinates of the probe.
k_nearest: Number of nearest points to return.
Returns:
dict with nearest_texts, nearest_indices, cluster_distances,
axis_percentages, etc.
"""
pts = result.projected_3d
probe = np.array([x, y, z])
# Distances to all points
dists = np.linalg.norm(pts - probe, axis=1)
nearest_indices = np.argsort(dists)[:k_nearest]
nearest_texts = [
{
"index": int(idx),
"text": result.session.entries[idx].text,
"role": result.session.entries[idx].role,
"distance": float(dists[idx]),
"cluster": result.clusters.labels[idx],
}
for idx in nearest_indices
]
# Axis percentages (0-100%)
mins = pts.min(axis=0)
maxs = pts.max(axis=0)
ranges = maxs - mins
ranges[ranges == 0] = 1.0
pct = ((probe - mins) / ranges) * 100
# Cluster distances (closeness %)
cluster_distances = {}
max_dist = float(np.linalg.norm(maxs - mins))
for c in range(result.clusters.n_clusters):
centroid = result.cluster_centroids_3d[c]
dist = float(np.linalg.norm(probe - centroid))
closeness = max(0, (1 - dist / max_dist) * 100) if max_dist > 0 else 0
label = (
result.cluster_labels[c]
if c < len(result.cluster_labels)
else f"Cluster {c}"
)
cluster_distances[label] = round(closeness, 1)
return {
"probe": {"x": x, "y": y, "z": z},
"axis_percentages": {
result.axis_info.labels[i]: round(float(pct[i]), 1) for i in range(3)
},
"nearest_texts": nearest_texts,
"cluster_distances": cluster_distances,
}
def probe_with_explanation(
result: AnalysisResult,
explainer: SemanticExplainer,
x: float,
y: float,
z: float,
k_nearest: int = 5,
) -> dict:
"""Probe a point and get an LLM explanation of its meaning.
Returns the same dict as probe_point plus an 'explanation' key.
"""
info = probe_point(result, x, y, z, k_nearest)
# Build cluster distances string
dist_str = ", ".join(
f"{name}: {pct}%" for name, pct in info["cluster_distances"].items()
)
nearest_texts = [item["text"] for item in info["nearest_texts"]]
pcts = info["axis_percentages"]
axis_labels = list(pcts.keys())
axis_vals = list(pcts.values())
explanation = explainer.explain_probe(
axis_labels=axis_labels,
x_val=axis_vals[0],
y_val=axis_vals[1],
z_val=axis_vals[2],
nearest_texts=nearest_texts,
cluster_distances=dist_str,
)
info["explanation"] = explanation
return info
|