Spaces:
Running on Zero
Running on Zero
File size: 17,683 Bytes
4791c0a 04ad98e 4791c0a 13fe947 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 13fe947 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a 04ad98e 4791c0a | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 | from __future__ import annotations
from collections import Counter, defaultdict
from collections.abc import Mapping, Sequence
import math
from typing import Any
from hackathon_advisor.data import (
Project,
ProjectIndex,
normalize_project_tags,
public_project_summary,
public_project_title,
tokenize,
)
from hackathon_advisor.quest_taxonomy import QUESTS, normalize_match, quest_profiles
from hackathon_advisor._text import utc_now
DASHBOARD_SCHEMA_VERSION = 1
TSNE_RANDOM_STATE = 42
TSNE_MIN_PROJECTS = 3
LINKS_PER_PROJECT = 2
CLUSTER_LABEL_ALGORITHM = "distinctive-keywords-v1"
STOPWORDS = {
"about",
"agent",
"app",
"apps",
"ai",
"all",
"an",
"and",
"are",
"as",
"at",
"before",
"assistant",
"be",
"been",
"being",
"build",
"build-small",
"build-small-hackathon",
"built",
"by",
"demo",
"face",
"for",
"from",
"gradio",
"hackathon",
"hugging",
"huggingface",
"in",
"is",
"it",
"its",
"first",
"local",
"make",
"makes",
"made",
"me",
"model",
"models",
"my",
"of",
"on",
"or",
"our",
"one",
"project",
"projects",
"pro",
"region",
"run",
"runs",
"small",
"space",
"spaces",
"submission",
"the",
"their",
"them",
"these",
"they",
"this",
"those",
"to",
"tool",
"tools",
"try",
"us",
"use",
"used",
"uses",
"using",
"we",
"with",
"you",
"your",
}
class DashboardError(ValueError):
pass
def build_dashboard_payload(
index: ProjectIndex,
*,
quest_matches: Mapping[str, Sequence[Mapping[str, Any]]] | None = None,
quest_source: str = "",
generated_at: str | None = None,
) -> dict[str, Any]:
projects = list(index.projects)
if len(projects) < TSNE_MIN_PROJECTS:
raise DashboardError(f"dashboard atlas requires at least {TSNE_MIN_PROJECTS} projects")
matrix = _embedding_matrix(index)
coordinates = _tsne_coordinates(matrix)
raw_cluster_labels = _cluster_labels(matrix)
cluster_id_by_raw, clusters = _cluster_payloads(projects, coordinates, raw_cluster_labels)
normalized_quest_matches = _normalize_quest_matches(projects, quest_matches)
points = _point_payloads(projects, coordinates, raw_cluster_labels, cluster_id_by_raw, normalized_quest_matches)
links = _nearest_links(projects, matrix)
quest_report = _quest_report(points, normalized_quest_matches, quest_source)
payload = {
"schema_version": DASHBOARD_SCHEMA_VERSION,
"generated_at": generated_at or utc_now(),
"project_count": len(projects),
"provenance": {
"snapshot_generated_at": index.generated_at,
"snapshot_source": index.source,
"index_generated_at": index.index_generated_at,
"index_algorithm": index.index_algorithm,
"snapshot_digest": index.snapshot_digest,
"embedding": index.embedding_metadata,
},
"layout": {
"algorithm": "tsne",
"metric": "cosine",
"init": "pca",
"random_state": TSNE_RANDOM_STATE,
"perplexity": _tsne_perplexity(len(projects)),
},
"cluster_label_algorithm": CLUSTER_LABEL_ALGORITHM,
"points": points,
"links": links,
"clusters": clusters,
"quest_report": quest_report,
}
validate_dashboard_payload(payload)
return payload
def validate_dashboard_payload(payload: Mapping[str, Any]) -> None:
if payload.get("schema_version") != DASHBOARD_SCHEMA_VERSION:
raise DashboardError("unsupported dashboard schema version")
project_count = int(payload.get("project_count") or 0)
if project_count < TSNE_MIN_PROJECTS:
raise DashboardError("dashboard project count is too small")
points = payload.get("points")
if not isinstance(points, list) or len(points) != project_count:
raise DashboardError("dashboard point count does not match project count")
ids: set[str] = set()
cluster_ids: set[str] = set()
for point in points:
if not isinstance(point, dict):
raise DashboardError("dashboard points must be objects")
project_id = str(point.get("id") or "")
if not project_id or project_id in ids:
raise DashboardError("dashboard points must have unique project ids")
ids.add(project_id)
x = float(point.get("x"))
y = float(point.get("y"))
if not 0.0 <= x <= 100.0 or not 0.0 <= y <= 100.0:
raise DashboardError("dashboard point coordinates must be percentages")
cluster_id = str(point.get("cluster_id") or "")
if not cluster_id:
raise DashboardError("dashboard point cluster id is missing")
cluster_ids.add(cluster_id)
clusters = payload.get("clusters")
if not isinstance(clusters, list) or not clusters:
raise DashboardError("dashboard clusters are missing")
declared_cluster_ids = {str(cluster.get("id") or "") for cluster in clusters if isinstance(cluster, dict)}
if cluster_ids - declared_cluster_ids:
raise DashboardError("dashboard points reference missing clusters")
links = payload.get("links")
if not isinstance(links, list):
raise DashboardError("dashboard links must be a list")
for link in links:
if str(link.get("source") or "") not in ids or str(link.get("target") or "") not in ids:
raise DashboardError("dashboard link references an unknown project")
quest_report = payload.get("quest_report")
if not isinstance(quest_report, dict):
raise DashboardError("dashboard quest report is missing")
if quest_report.get("status") not in {"analyzed", "not_analyzed"}:
raise DashboardError("dashboard quest report status is invalid")
def _embedding_matrix(index: ProjectIndex) -> Any:
import numpy as np
return np.asarray(index.project_vectors(), dtype=np.float32)
def _tsne_coordinates(matrix: Any) -> list[tuple[float, float]]:
from sklearn.manifold import TSNE
coords = TSNE(
n_components=2,
perplexity=_tsne_perplexity(int(matrix.shape[0])),
init="pca",
learning_rate="auto",
max_iter=1000,
metric="cosine",
random_state=TSNE_RANDOM_STATE,
).fit_transform(matrix)
return _scale_points(coords)
def _tsne_perplexity(count: int) -> int:
return max(2, min(30, count // 4))
def _cluster_labels(matrix: Any) -> list[int]:
from sklearn.cluster import KMeans
count = int(matrix.shape[0])
cluster_count = min(10, max(min(6, count), round(math.sqrt(count))))
labels = KMeans(
n_clusters=cluster_count,
random_state=TSNE_RANDOM_STATE,
n_init=20,
).fit_predict(matrix)
return [int(label) for label in labels]
def _scale_points(points: Any, low: float = 3.0, high: float = 97.0) -> list[tuple[float, float]]:
import numpy as np
scaled = np.empty_like(points, dtype=np.float64)
for axis in range(points.shape[1]):
column = points[:, axis]
minimum = float(column.min())
maximum = float(column.max())
span = maximum - minimum
if span <= 1e-9:
scaled[:, axis] = (low + high) / 2.0
else:
scaled[:, axis] = low + (column - minimum) / span * (high - low)
return [(round(float(x), 4), round(float(y), 4)) for x, y in scaled]
def _cluster_payloads(
projects: Sequence[Project],
coordinates: Sequence[tuple[float, float]],
raw_labels: Sequence[int],
) -> tuple[dict[int, str], list[dict[str, Any]]]:
grouped: dict[int, list[int]] = defaultdict(list)
for index, label in enumerate(raw_labels):
grouped[int(label)].append(index)
ordered_raw_labels = sorted(
grouped,
key=lambda label: (-len(grouped[label]), _cluster_center(coordinates, grouped[label])),
)
cluster_id_by_raw = {label: f"cluster-{position + 1}" for position, label in enumerate(ordered_raw_labels)}
clusters: list[dict[str, Any]] = []
corpus_document_frequency = _corpus_document_frequency(projects)
for raw_label in ordered_raw_labels:
indexes = grouped[raw_label]
cluster_projects = [projects[index] for index in indexes]
representatives = sorted(
cluster_projects,
key=lambda project: (project.likes, project.last_modified, project.title.lower()),
reverse=True,
)[:4]
keywords = _cluster_keywords(
cluster_projects,
corpus_document_frequency=corpus_document_frequency,
corpus_project_count=len(projects),
)
label = (
" / ".join(word.title() for word in keywords[:2])
if keywords
else _representative_cluster_label(representatives)
)
clusters.append(
{
"id": cluster_id_by_raw[raw_label],
"label": label,
"keywords": keywords,
"project_count": len(indexes),
"center": {
"x": round(sum(coordinates[index][0] for index in indexes) / len(indexes), 4),
"y": round(sum(coordinates[index][1] for index in indexes) / len(indexes), 4),
},
"representative_projects": [project.to_public_dict() for project in representatives],
}
)
return cluster_id_by_raw, clusters
def _cluster_center(coordinates: Sequence[tuple[float, float]], indexes: Sequence[int]) -> tuple[float, float]:
return (
sum(coordinates[index][0] for index in indexes) / len(indexes),
sum(coordinates[index][1] for index in indexes) / len(indexes),
)
def _corpus_document_frequency(projects: Sequence[Project]) -> Counter[str]:
document_frequency: Counter[str] = Counter()
for project in projects:
document_frequency.update(set(_project_keyword_tokens(project)))
return document_frequency
def _cluster_keywords(
projects: Sequence[Project],
*,
corpus_document_frequency: Mapping[str, int],
corpus_project_count: int,
) -> list[str]:
counts: Counter[str] = Counter()
document_frequency: Counter[str] = Counter()
project_list = list(projects)
for project in project_list:
tokens = _project_keyword_tokens(project)
counts.update(tokens)
document_frequency.update(set(tokens))
if not project_list:
return []
min_cluster_documents = 1 if len(project_list) <= 3 else 2
scored: list[tuple[float, int, int, str]] = []
for token, count in counts.items():
cluster_documents = document_frequency[token]
if cluster_documents < min_cluster_documents:
continue
corpus_documents = int(corpus_document_frequency.get(token) or 0)
if corpus_documents <= 0:
continue
inverse_document_frequency = math.log((1 + corpus_project_count) / (1 + corpus_documents))
if inverse_document_frequency <= 0.0:
continue
exclusivity = cluster_documents / corpus_documents
coverage = cluster_documents / len(project_list)
score = (
(1.0 + math.log(count))
* inverse_document_frequency
* (0.35 + 0.65 * exclusivity)
* (0.35 + 0.65 * coverage)
)
scored.append((score, cluster_documents, count, token))
scored.sort(key=lambda item: (-item[0], -item[1], -item[2], item[3]))
return [token for _score, _cluster_documents, _count, token in scored[:5]]
def _project_keyword_tokens(project: Project) -> list[str]:
text = " ".join(
[
project.title,
project.slug.replace("-", " ").replace("_", " "),
project.summary,
" ".join(normalize_project_tags(project.tags)),
" ".join(project.models),
]
)
return [token for token in tokenize(text) if _is_cluster_keyword(token)]
def _is_cluster_keyword(token: str) -> bool:
if token in STOPWORDS:
return False
if token.startswith("region"):
return False
if token.isdigit():
return False
return True
def _representative_cluster_label(projects: Sequence[Project]) -> str:
labels: list[str] = []
for project in projects:
title = public_project_title(project.title)
if title == "Untitled project":
continue
labels.append(title)
if len(labels) == 2:
break
return " / ".join(labels) if labels else "Mixed projects"
def _normalize_quest_matches(
projects: Sequence[Project],
quest_matches: Mapping[str, Sequence[Mapping[str, Any]]] | None,
) -> dict[str, list[dict[str, Any]]]:
project_ids = {project.id for project in projects}
normalized = {project.id: [] for project in projects}
if quest_matches is None:
return normalized
if set(quest_matches) != project_ids:
missing = sorted(project_ids - set(quest_matches))
extra = sorted(set(quest_matches) - project_ids)
detail = []
if missing:
detail.append(f"missing {len(missing)} projects")
if extra:
detail.append(f"unknown {len(extra)} projects")
raise DashboardError("quest analysis project coverage is invalid: " + ", ".join(detail))
for project_id, matches in quest_matches.items():
normalized[project_id] = [_normalize_quest_match(match) for match in matches]
return normalized
def _normalize_quest_match(match: Mapping[str, Any]) -> dict[str, Any]:
try:
return normalize_match(match)
except ValueError as error:
raise DashboardError(f"invalid quest match: {error}") from error
def _point_payloads(
projects: Sequence[Project],
coordinates: Sequence[tuple[float, float]],
raw_labels: Sequence[int],
cluster_id_by_raw: Mapping[int, str],
quest_matches: Mapping[str, Sequence[Mapping[str, Any]]],
) -> list[dict[str, Any]]:
points: list[dict[str, Any]] = []
for project, (x, y), raw_label in zip(projects, coordinates, raw_labels, strict=True):
matches = list(quest_matches.get(project.id) or [])
points.append(
{
"id": project.id,
"title": public_project_title(project.title),
"summary": public_project_summary(project.summary),
"url": project.url,
"host": project.host,
"likes": project.likes,
"sdk": project.sdk,
"models": list(project.models),
"tags": list(normalize_project_tags(project.tags)),
"last_modified": project.last_modified,
"x": x,
"y": y,
"cluster_id": cluster_id_by_raw[int(raw_label)],
"quest_matches": matches,
"quest_ids": [str(match["quest"]) for match in matches],
}
)
return points
def _nearest_links(projects: Sequence[Project], matrix: Any) -> list[dict[str, Any]]:
import numpy as np
similarity = matrix @ matrix.T
pairs: dict[tuple[int, int], float] = {}
for index in range(len(projects)):
order = np.argsort(similarity[index])[::-1]
neighbors = [int(candidate) for candidate in order if int(candidate) != index][:LINKS_PER_PROJECT]
for neighbor in neighbors:
left, right = sorted((index, neighbor))
pairs[(left, right)] = max(float(similarity[left, right]), pairs.get((left, right), -1.0))
return [
{
"source": projects[left].id,
"target": projects[right].id,
"score": round(max(0.0, min(1.0, score)), 4),
}
for (left, right), score in sorted(pairs.items(), key=lambda item: (-item[1], item[0]))
]
def _quest_report(
points: Sequence[Mapping[str, Any]],
quest_matches: Mapping[str, Sequence[Mapping[str, Any]]],
quest_source: str,
) -> dict[str, Any]:
profiles = {profile["id"]: profile for profile in quest_profiles()}
status = "analyzed" if quest_source else "not_analyzed"
quests = []
for quest in QUESTS:
matched_points = [
point
for point in points
if any(match["quest"] == quest for match in quest_matches.get(str(point["id"]), []))
]
examples = sorted(
matched_points,
key=lambda point: (
max(
(
float(match["confidence"])
for match in quest_matches.get(str(point["id"]), [])
if match["quest"] == quest
),
default=0.0,
),
int(point.get("likes") or 0),
),
reverse=True,
)[:4]
profile = profiles.get(quest, {"label": quest, "description": ""})
quests.append(
{
"id": quest,
"label": profile["label"],
"description": profile["description"],
"project_count": len(matched_points),
"examples": [
{
"id": point["id"],
"title": point["title"],
"url": point["url"],
}
for point in examples
],
}
)
return {
"status": status,
"source": quest_source,
"quests": quests,
}
|