Spaces:
Running
Running
Commit Β·
e4a4c32
1
Parent(s): 75ab97e
cot
Browse files
app.py
CHANGED
|
@@ -55,7 +55,6 @@ cot_by_agent = {}
|
|
| 55 |
if os.path.exists(COT_PATH):
|
| 56 |
with open(COT_PATH, "r") as f:
|
| 57 |
cot_raw = json.load(f)
|
| 58 |
-
# Support both list and {"inference_results": [...]} formats
|
| 59 |
records = cot_raw if isinstance(cot_raw, list) else cot_raw.get("inference_results", [])
|
| 60 |
for result in records:
|
| 61 |
cot_by_agent[int(result["agent_id"])] = result
|
|
@@ -70,7 +69,10 @@ def get_cot(agent_id):
|
|
| 70 |
s1 = result.get("step1_response", "")
|
| 71 |
s2 = result.get("step2_response", "")
|
| 72 |
s3 = result.get("step3_response", "")
|
| 73 |
-
|
|
|
|
|
|
|
|
|
|
| 74 |
|
| 75 |
|
| 76 |
# ββ Mobility text builders ββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
|
@@ -85,11 +87,9 @@ def build_mobility_summary(agent_sp):
|
|
| 85 |
obs_end = agent_sp["end_datetime"].max().strftime("%Y-%m-%d")
|
| 86 |
days = (agent_sp["end_datetime"].max() - agent_sp["start_datetime"].min()).days
|
| 87 |
|
| 88 |
-
# Top activity types
|
| 89 |
act_counts = agent_sp["act_label"].value_counts().head(3)
|
| 90 |
top_acts = ", ".join(f"{a} ({n})" for a, n in act_counts.items())
|
| 91 |
|
| 92 |
-
# Time of day
|
| 93 |
agent_sp2 = agent_sp.copy()
|
| 94 |
agent_sp2["hour"] = agent_sp2["start_datetime"].dt.hour
|
| 95 |
def tod(h):
|
|
@@ -144,10 +144,6 @@ def build_weekly_checkin(agent_sp, max_days=None):
|
|
| 144 |
|
| 145 |
# ββ HTML reasoning chain ββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 146 |
|
| 147 |
-
# ββ Paste this entire block into app.py, replacing the existing CHAIN_CSS, render_chain, and helper functions ββ
|
| 148 |
-
|
| 149 |
-
import re
|
| 150 |
-
|
| 151 |
CHAIN_CSS = """
|
| 152 |
<style>
|
| 153 |
@import url('https://fonts.googleapis.com/css2?family=DM+Mono:wght@400;500&family=DM+Sans:wght@300;400;500;600&display=swap');
|
|
@@ -166,7 +162,7 @@ CHAIN_CSS = """
|
|
| 166 |
overflow: hidden;
|
| 167 |
transition: opacity 0.3s, filter 0.3s;
|
| 168 |
}
|
| 169 |
-
.hct-stage.dim
|
| 170 |
.hct-stage.active { opacity: 1; }
|
| 171 |
|
| 172 |
/* ββ Stage header strip ββ */
|
|
@@ -189,270 +185,270 @@ CHAIN_CSS = """
|
|
| 189 |
font-weight: 600;
|
| 190 |
letter-spacing: 0.04em;
|
| 191 |
text-transform: uppercase;
|
|
|
|
| 192 |
}
|
| 193 |
|
| 194 |
/* Stage 1 colors */
|
| 195 |
.hct-s1 { background: #f4f6fb; border: 1.5px solid #d4daf0; }
|
| 196 |
.hct-s1 .hct-head { background: #eaecf7; border-bottom: 1px solid #d4daf0; }
|
| 197 |
-
.hct-s1 .hct-num
|
| 198 |
.hct-s1 .hct-title { color: #3a4a80; }
|
| 199 |
|
| 200 |
/* Stage 2 colors */
|
| 201 |
.hct-s2 { background: #fdf8f2; border: 1.5px solid #e8d5b8; }
|
| 202 |
.hct-s2 .hct-head { background: #f7ede0; border-bottom: 1px solid #e8d5b8; }
|
| 203 |
-
.hct-s2 .hct-num
|
| 204 |
.hct-s2 .hct-title { color: #7a4a10; }
|
| 205 |
|
| 206 |
/* Stage 3 colors */
|
| 207 |
.hct-s3 { background: #fff6f5; border: 2px solid #d4453a; }
|
| 208 |
.hct-s3 .hct-head { background: #fce8e7; border-bottom: 1px solid #d4453a; }
|
| 209 |
-
.hct-s3 .hct-num
|
| 210 |
.hct-s3 .hct-title { color: #b0302a; }
|
| 211 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 212 |
/* ββ Body ββ */
|
| 213 |
.hct-body { padding: 12px 14px; }
|
| 214 |
|
| 215 |
/* ββ Arrow connector ββ */
|
| 216 |
.hct-arrow {
|
| 217 |
-
display: flex;
|
| 218 |
-
|
| 219 |
-
gap: 8px;
|
| 220 |
-
padding: 5px 18px;
|
| 221 |
-
transition: opacity 0.3s;
|
| 222 |
}
|
| 223 |
-
.hct-arrow-line
|
| 224 |
.hct-arrow-label {
|
| 225 |
-
font-family: 'DM Mono', monospace;
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
| 229 |
-
text-transform: uppercase;
|
| 230 |
-
white-space: nowrap;
|
| 231 |
-
background: white;
|
| 232 |
-
padding: 2px 8px;
|
| 233 |
-
border: 1px solid #e0dbd4;
|
| 234 |
-
border-radius: 20px;
|
| 235 |
}
|
| 236 |
|
| 237 |
/* ββ Stage 1: Location table ββ */
|
| 238 |
.hct-loc-table {
|
| 239 |
-
width: 100%;
|
| 240 |
-
|
| 241 |
-
font-size: 11.5px;
|
| 242 |
-
margin-bottom: 10px;
|
| 243 |
}
|
| 244 |
.hct-loc-table th {
|
| 245 |
-
font-family: 'DM Mono', monospace;
|
| 246 |
-
|
| 247 |
-
|
| 248 |
-
letter-spacing: 0.1em;
|
| 249 |
-
text-transform: uppercase;
|
| 250 |
-
color: #8090b0;
|
| 251 |
-
border-bottom: 1px solid #d4daf0;
|
| 252 |
-
padding: 3px 6px 5px;
|
| 253 |
-
text-align: left;
|
| 254 |
}
|
| 255 |
.hct-loc-table th:not(:first-child) { text-align: right; }
|
| 256 |
.hct-loc-table td {
|
| 257 |
-
padding: 5px 6px;
|
| 258 |
-
|
| 259 |
-
|
| 260 |
-
|
|
|
|
|
|
|
| 261 |
}
|
| 262 |
-
.hct-loc-table td:not(:first-child) { text-align: right; font-family: 'DM Mono', monospace; font-size: 11px; color: #5060a0; }
|
| 263 |
.hct-loc-table tr:last-child td { border-bottom: none; }
|
| 264 |
-
.hct-loc-name {
|
|
|
|
|
|
|
|
|
|
| 265 |
.hct-visit-bar-wrap { display: flex; align-items: center; gap: 6px; justify-content: flex-end; }
|
| 266 |
.hct-visit-bar { height: 4px; border-radius: 2px; background: #6878c8; opacity: 0.55; }
|
| 267 |
|
| 268 |
/* ββ Stage 1: Temporal panel ββ */
|
| 269 |
-
.hct-temporal {
|
| 270 |
-
|
| 271 |
-
grid-template-columns: 1fr 1fr;
|
| 272 |
-
gap: 8px;
|
| 273 |
-
}
|
| 274 |
-
.hct-temp-block {
|
| 275 |
-
background: #eef0fa;
|
| 276 |
-
border-radius: 8px;
|
| 277 |
-
padding: 8px 10px;
|
| 278 |
-
}
|
| 279 |
.hct-temp-label {
|
| 280 |
-
font-family: 'DM Mono', monospace;
|
| 281 |
-
|
| 282 |
-
font-weight: 500;
|
| 283 |
-
letter-spacing: 0.1em;
|
| 284 |
-
text-transform: uppercase;
|
| 285 |
-
color: #7080b0;
|
| 286 |
-
margin-bottom: 6px;
|
| 287 |
}
|
| 288 |
.hct-seg-row { display: flex; height: 10px; border-radius: 5px; overflow: hidden; margin-bottom: 5px; }
|
| 289 |
-
.hct-seg {
|
| 290 |
-
.seg-morning
|
| 291 |
-
.seg-afternoon{ background: #f97316; }
|
| 292 |
-
.seg-evening
|
| 293 |
-
.seg-night
|
| 294 |
-
.seg-weekday
|
| 295 |
-
.seg-weekend
|
| 296 |
.hct-legend { display: flex; flex-wrap: wrap; gap: 4px 10px; }
|
| 297 |
.hct-leg-item { display: flex; align-items: center; gap: 4px; font-size: 10px; color: #5a6080; }
|
| 298 |
.hct-leg-dot { width: 8px; height: 8px; border-radius: 2px; flex-shrink: 0; }
|
| 299 |
.hct-dist-line {
|
| 300 |
-
margin-top: 8px;
|
| 301 |
-
font-
|
| 302 |
-
|
| 303 |
-
|
| 304 |
-
padding: 5px 8px;
|
| 305 |
-
background: #eef0fa;
|
| 306 |
-
border-radius: 6px;
|
| 307 |
-
display: flex;
|
| 308 |
-
align-items: center;
|
| 309 |
-
gap: 6px;
|
| 310 |
}
|
| 311 |
|
| 312 |
-
/* ββ Stage 2:
|
| 313 |
-
.hct-dim-grid {
|
| 314 |
-
display: grid;
|
| 315 |
-
grid-template-columns: 1fr 1fr;
|
| 316 |
-
gap: 8px;
|
| 317 |
-
}
|
| 318 |
.hct-dim-card {
|
| 319 |
-
background: #fff;
|
| 320 |
-
border:
|
| 321 |
-
border-radius: 8px;
|
| 322 |
-
padding: 9px 11px;
|
| 323 |
-
}
|
| 324 |
-
.hct-dim-head {
|
| 325 |
-
display: flex;
|
| 326 |
-
align-items: center;
|
| 327 |
-
gap: 6px;
|
| 328 |
-
margin-bottom: 5px;
|
| 329 |
}
|
|
|
|
| 330 |
.hct-dim-icon { font-size: 13px; line-height: 1; }
|
| 331 |
.hct-dim-name {
|
| 332 |
-
font-family: 'DM Mono', monospace;
|
| 333 |
-
|
| 334 |
-
font-weight: 500;
|
| 335 |
-
letter-spacing: 0.1em;
|
| 336 |
-
text-transform: uppercase;
|
| 337 |
-
color: #a07040;
|
| 338 |
-
}
|
| 339 |
-
.hct-dim-text {
|
| 340 |
-
font-size: 11px;
|
| 341 |
-
color: #3a2a10;
|
| 342 |
-
line-height: 1.55;
|
| 343 |
}
|
|
|
|
| 344 |
.hct-dim-empty { color: #ccc; font-style: italic; }
|
| 345 |
|
| 346 |
-
/* ββ Stage 3
|
| 347 |
-
.hct-pred-row {
|
| 348 |
-
display: flex;
|
| 349 |
-
align-items: flex-start;
|
| 350 |
-
gap: 16px;
|
| 351 |
-
margin-bottom: 10px;
|
| 352 |
-
}
|
| 353 |
.hct-pred-badge {
|
| 354 |
-
background: #d4453a;
|
| 355 |
-
|
| 356 |
-
border-radius: 8px;
|
| 357 |
-
padding: 8px 14px;
|
| 358 |
-
text-align: center;
|
| 359 |
-
flex-shrink: 0;
|
| 360 |
}
|
| 361 |
.hct-pred-val { font-size: 18px; font-weight: 600; line-height: 1.2; white-space: nowrap; }
|
| 362 |
-
.hct-pred-sub {
|
| 363 |
-
|
| 364 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 365 |
.hct-conf-track { height: 6px; background: #f0d0cf; border-radius: 3px; overflow: hidden; margin-bottom: 6px; }
|
| 366 |
.hct-conf-fill { height: 100%; background: linear-gradient(90deg, #e74c3c, #8b0000); border-radius: 3px; }
|
| 367 |
.hct-reasoning {
|
| 368 |
-
font-size: 11.5px;
|
| 369 |
-
|
| 370 |
-
line-height: 1.6;
|
| 371 |
-
border-left: 3px solid #e8b0ae;
|
| 372 |
-
padding-left: 10px;
|
| 373 |
}
|
| 374 |
|
| 375 |
-
/* ββ Idle / loading
|
| 376 |
.hct-idle { font-size: 12px; color: #b0bac8; padding: 6px 0; font-style: italic; }
|
| 377 |
-
.hct-loading {
|
| 378 |
-
|
| 379 |
-
|
|
|
|
| 380 |
}
|
| 381 |
-
.hct-dot { width: 6px; height: 6px; border-radius: 50%; display: inline-block; animation: hct-pulse 1.2s ease-in-out infinite; }
|
| 382 |
.hct-dot:nth-child(2) { animation-delay: 0.2s; }
|
| 383 |
.hct-dot:nth-child(3) { animation-delay: 0.4s; }
|
| 384 |
@keyframes hct-pulse {
|
| 385 |
0%,100% { opacity: 0.2; transform: scale(0.8); }
|
| 386 |
-
50%
|
| 387 |
}
|
| 388 |
.hct-s1 .hct-dot { background: #6878c8; }
|
| 389 |
.hct-s2 .hct-dot { background: #c08040; }
|
| 390 |
.hct-s3 .hct-dot { background: #d4453a; }
|
| 391 |
</style>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 392 |
"""
|
| 393 |
|
| 394 |
|
| 395 |
def _loading(msg):
|
| 396 |
-
return f'<div class="hct-loading">
|
|
|
|
|
|
|
| 397 |
|
| 398 |
|
|
|
|
|
|
|
| 399 |
def _parse_s1(text):
|
| 400 |
-
|
| 401 |
-
locations = []
|
| 402 |
-
dur_map = {}
|
| 403 |
-
tod = {}
|
| 404 |
-
wk = {}
|
| 405 |
-
dist = None
|
| 406 |
|
| 407 |
for line in text.splitlines():
|
| 408 |
s = line.strip()
|
| 409 |
-
|
| 410 |
-
|
|
|
|
| 411 |
if m:
|
| 412 |
locations.append((m.group(1).strip(), int(m.group(2))))
|
| 413 |
-
|
| 414 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 415 |
if m2:
|
| 416 |
dur_map[m2.group(1).strip()] = float(m2.group(2))
|
| 417 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 418 |
if not tod:
|
| 419 |
-
|
| 420 |
-
if
|
| 421 |
-
|
| 422 |
-
|
|
|
|
|
|
|
|
|
|
| 423 |
# Weekday/weekend
|
| 424 |
if not wk:
|
| 425 |
m4 = re.search(r'(\d+)%\s*weekday.*?(\d+)%\s*weekend', s, re.IGNORECASE)
|
| 426 |
if m4:
|
| 427 |
wk = {'Weekday': int(m4.group(1)), 'Weekend': int(m4.group(2))}
|
|
|
|
| 428 |
# Distance
|
| 429 |
if not dist:
|
| 430 |
-
m5 = re.search(r'average distance of approximately ([\d.]+)\s*miles', s, re.IGNORECASE)
|
| 431 |
if m5:
|
| 432 |
dist = float(m5.group(1))
|
| 433 |
|
| 434 |
-
|
| 435 |
-
return result_locs, tod, wk, dist
|
| 436 |
|
| 437 |
|
| 438 |
def _parse_s2(text):
|
| 439 |
-
"""Returns dict: ROUTINE, ECONOMIC, SOCIAL, URBAN, STABILITY β short summary string"""
|
| 440 |
DIMS = {
|
| 441 |
-
'ROUTINE':
|
| 442 |
-
'ECONOMIC':
|
| 443 |
-
'SOCIAL':
|
| 444 |
-
'
|
| 445 |
-
'STABILITY': ['STABILITY', 'REGULARITY', 'CONSISTENCY'],
|
| 446 |
}
|
| 447 |
-
sections = {}
|
| 448 |
-
current_key = None
|
| 449 |
-
current_lines = []
|
| 450 |
|
| 451 |
for line in text.splitlines():
|
| 452 |
s = line.strip()
|
| 453 |
-
# Format A: "1. TITLE ANALYSIS:" or "2. ECONOMIC BEHAVIOR PATTERNS:"
|
| 454 |
mA = re.match(r'^\d+\.\s+([A-Z][A-Z\s&]+?)(?:\s+ANALYSIS|\s+PATTERNS|\s+INDICATORS|\s+CHARACTERISTICS|\s+STABILITY)?:\s*$', s, re.IGNORECASE)
|
| 455 |
-
# Format B: "STEP 1: ROUTINE & SCHEDULE ANALYSIS"
|
| 456 |
mB = re.match(r'^STEP\s+\d+:\s+([A-Z][A-Z\s&]+?)(?:\s+ANALYSIS|\s+PATTERNS|\s+INDICATORS|\s+CHARACTERISTICS|\s+STABILITY)?\s*$', s, re.IGNORECASE)
|
| 457 |
mm = mA or mB
|
| 458 |
if mm:
|
|
@@ -463,8 +459,7 @@ def _parse_s2(text):
|
|
| 463 |
elif current_key and s:
|
| 464 |
if re.match(r'^\d+\.\d+', s):
|
| 465 |
sub = re.sub(r'^\d+\.\d+[^:]*:\s*', '', s)
|
| 466 |
-
if sub:
|
| 467 |
-
current_lines.append(sub)
|
| 468 |
elif s.startswith('-'):
|
| 469 |
current_lines.append(s.lstrip('-').strip())
|
| 470 |
elif not re.match(r'^\d+\.', s):
|
|
@@ -479,105 +474,86 @@ def _parse_s2(text):
|
|
| 479 |
if any(kw in k for kw in keywords) and txt:
|
| 480 |
sents = re.split(r'(?<=[.!?])\s+', txt.strip())
|
| 481 |
summary = ' '.join(sents[:2])
|
| 482 |
-
if len(summary) > 160
|
| 483 |
-
summary = summary[:157] + 'β¦'
|
| 484 |
-
result[dim] = summary
|
| 485 |
break
|
| 486 |
return result
|
| 487 |
|
| 488 |
|
| 489 |
def _parse_s3(text):
|
| 490 |
-
pred, conf,
|
| 491 |
-
in_r = False
|
| 492 |
-
r_lines = []
|
| 493 |
for line in text.splitlines():
|
| 494 |
s = line.strip()
|
| 495 |
if s.startswith('INCOME_PREDICTION:'):
|
| 496 |
pred = s.replace('INCOME_PREDICTION:', '').strip()
|
| 497 |
elif s.startswith('INCOME_CONFIDENCE:'):
|
| 498 |
-
try:
|
| 499 |
-
|
| 500 |
-
except:
|
| 501 |
-
conf = 0
|
| 502 |
elif s.startswith('INCOME_REASONING:'):
|
| 503 |
in_r = True
|
| 504 |
r_lines.append(s.replace('INCOME_REASONING:', '').strip())
|
| 505 |
elif in_r:
|
| 506 |
-
if re.match(r'^2\.', s) or s.startswith('INCOME_'):
|
| 507 |
-
|
| 508 |
-
if s:
|
| 509 |
-
r_lines.append(s)
|
| 510 |
reasoning = ' '.join(r_lines).strip()
|
| 511 |
sents = re.split(r'(?<=[.!?])\s+', reasoning)
|
| 512 |
reasoning = ' '.join(sents[:3])
|
| 513 |
-
if len(reasoning) > 280
|
| 514 |
-
|
| 515 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 516 |
|
| 517 |
|
|
|
|
|
|
|
| 518 |
def _s1_body(text, active):
|
| 519 |
if not active:
|
| 520 |
return '<div class="hct-idle">Press βΆ to start</div>'
|
| 521 |
if not text:
|
| 522 |
return _loading('Extracting features')
|
|
|
|
| 523 |
locs, tod, wk, dist = _parse_s1(text)
|
| 524 |
|
| 525 |
-
# Location table
|
| 526 |
max_v = max((v for _, v, _ in locs), default=1)
|
| 527 |
rows = ''
|
| 528 |
for name, visits, dur in locs:
|
| 529 |
bar_w = int(60 * visits / max_v)
|
| 530 |
dur_str = f'{int(dur)}m' if dur else 'β'
|
| 531 |
-
rows += (
|
| 532 |
-
|
| 533 |
-
|
| 534 |
-
|
| 535 |
-
|
| 536 |
-
|
| 537 |
-
|
| 538 |
-
|
| 539 |
-
|
| 540 |
-
table = (
|
| 541 |
-
f'<table class="hct-loc-table">'
|
| 542 |
-
f'<thead><tr><th>Location</th><th>Visits</th><th>Avg Stay</th></tr></thead>'
|
| 543 |
-
f'<tbody>{rows}</tbody>'
|
| 544 |
-
f'</table>'
|
| 545 |
-
) if rows else ''
|
| 546 |
-
|
| 547 |
-
# Temporal panels
|
| 548 |
def seg_bar(data, seg_classes):
|
| 549 |
total = sum(data.values()) or 1
|
| 550 |
segs = ''.join(
|
| 551 |
f'<div class="hct-seg {cls}" style="width:{int(100*v/total)}%"></div>'
|
| 552 |
-
for (label, v), cls in zip(data.items(), seg_classes)
|
| 553 |
-
)
|
| 554 |
legend = ''.join(
|
| 555 |
f'<div class="hct-leg-item"><div class="hct-leg-dot {cls}"></div>{label} {v}%</div>'
|
| 556 |
-
for (label, v), cls in zip(data.items(), seg_classes)
|
| 557 |
-
)
|
| 558 |
return f'<div class="hct-seg-row">{segs}</div><div class="hct-legend">{legend}</div>'
|
| 559 |
|
| 560 |
-
tod_panel = ''
|
| 561 |
-
|
| 562 |
-
|
| 563 |
-
|
| 564 |
-
|
| 565 |
-
|
| 566 |
-
f'</div>'
|
| 567 |
-
)
|
| 568 |
-
wk_panel = ''
|
| 569 |
-
if wk:
|
| 570 |
-
wk_panel = (
|
| 571 |
-
f'<div class="hct-temp-block">'
|
| 572 |
-
f'<div class="hct-temp-label">Weekday / Weekend</div>'
|
| 573 |
-
f'{seg_bar(wk, ["seg-weekday","seg-weekend"])}'
|
| 574 |
-
f'</div>'
|
| 575 |
-
)
|
| 576 |
-
temporal = f'<div class="hct-temporal">{tod_panel}{wk_panel}</div>' if (tod_panel or wk_panel) else ''
|
| 577 |
-
|
| 578 |
-
dist_line = ''
|
| 579 |
-
if dist:
|
| 580 |
-
dist_line = f'<div class="hct-dist-line">π Avg trip distance {dist} mi</div>'
|
| 581 |
|
| 582 |
return table + temporal + dist_line
|
| 583 |
|
|
@@ -588,30 +564,18 @@ def _s2_body(text, active):
|
|
| 588 |
if not text:
|
| 589 |
return _loading('Analyzing behavior')
|
| 590 |
dims = _parse_s2(text)
|
| 591 |
-
|
| 592 |
-
|
| 593 |
-
('ROUTINE', 'π', 'Schedule'),
|
| 594 |
-
('ECONOMIC', 'π°', 'Economic'),
|
| 595 |
-
('SOCIAL', 'π₯', 'Social'),
|
| 596 |
-
('STABILITY', 'π', 'Stability'),
|
| 597 |
-
]
|
| 598 |
-
# fallback to URBAN if STABILITY missing
|
| 599 |
-
if 'STABILITY' not in dims and 'URBAN' in dims:
|
| 600 |
-
dims['STABILITY'] = dims['URBAN']
|
| 601 |
-
|
| 602 |
cards = ''
|
| 603 |
for key, icon, label in DIM_META:
|
| 604 |
txt = dims.get(key, '')
|
| 605 |
-
content = f'<div class="hct-dim-text">{txt}</div>' if txt
|
| 606 |
-
|
| 607 |
-
|
| 608 |
-
|
| 609 |
-
|
| 610 |
-
|
| 611 |
-
|
| 612 |
-
f'{content}'
|
| 613 |
-
f'</div>'
|
| 614 |
-
)
|
| 615 |
return f'<div class="hct-dim-grid">{cards}</div>'
|
| 616 |
|
| 617 |
|
|
@@ -621,60 +585,63 @@ def _s3_body(text, active):
|
|
| 621 |
if not text:
|
| 622 |
return _loading('Inferring demographics')
|
| 623 |
pred, conf, reasoning = _parse_s3(text)
|
| 624 |
-
|
| 625 |
-
|
| 626 |
-
|
| 627 |
-
|
| 628 |
-
|
| 629 |
-
|
| 630 |
-
|
| 631 |
-
f'<div class="hct-conf-col">'
|
| 632 |
-
f'<div class="hct-conf-label">Confidence {conf}/5</div>'
|
| 633 |
-
f'<div class="hct-conf-track"><div class="hct-conf-fill" style="width:{conf_pct}%"></div></div>'
|
| 634 |
-
f'</div>'
|
| 635 |
-
f'</div>'
|
| 636 |
-
f'<div class="hct-reasoning">{reasoning}</div>'
|
| 637 |
-
)
|
| 638 |
|
|
|
|
| 639 |
|
| 640 |
-
def render_chain(s1_text, s2_text, s3_text, status="idle"
|
|
|
|
| 641 |
s1_on = status in ("running1", "running2", "running3", "done")
|
| 642 |
s2_on = status in ("running2", "running3", "done")
|
| 643 |
s3_on = status in ("running3", "done")
|
| 644 |
|
| 645 |
-
# For "running" states the text may be empty β show loading dots
|
| 646 |
s1_body = _s1_body(s1_text if s1_on else '', s1_on)
|
| 647 |
s2_body = _s2_body(s2_text if s2_on else '', s2_on)
|
| 648 |
s3_body = _s3_body(s3_text if s3_on else '', s3_on)
|
| 649 |
|
| 650 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 651 |
dim_cls = 'active' if on else 'dim'
|
| 652 |
-
|
| 653 |
-
|
| 654 |
-
|
| 655 |
-
|
| 656 |
-
|
| 657 |
-
|
| 658 |
-
|
| 659 |
-
|
| 660 |
-
|
| 661 |
|
| 662 |
def arrow(label, on):
|
| 663 |
op = '1' if on else '0.2'
|
| 664 |
-
return (
|
| 665 |
-
|
| 666 |
-
|
| 667 |
-
|
| 668 |
-
f'<div class="hct-arrow-line"></div>'
|
| 669 |
-
f'</div>'
|
| 670 |
-
)
|
| 671 |
|
| 672 |
-
html
|
| 673 |
-
html += stage('s1', 'Stage 01', 'Feature Extraction',
|
| 674 |
html += arrow('behavioral abstraction', s2_on)
|
| 675 |
-
html += stage('s2', 'Stage 02', 'Behavioral Analysis',
|
| 676 |
html += arrow('demographic inference', s3_on)
|
| 677 |
-
html += stage('s3', 'Stage 03', 'Demographic Inference', s3_body, s3_on)
|
| 678 |
html += '</div>'
|
| 679 |
return html
|
| 680 |
|
|
@@ -768,20 +735,17 @@ def on_select(agent_id):
|
|
| 768 |
|
| 769 |
|
| 770 |
def run_step(agent_id, step):
|
| 771 |
-
"""Reveal one stage per click. step: 0->1->2->done(-1)"""
|
| 772 |
agent_id = int(agent_id)
|
| 773 |
-
s1, s2, s3 = get_cot(agent_id)
|
| 774 |
next_step = step + 1
|
| 775 |
if next_step == 1:
|
| 776 |
-
html = render_chain(s1, "", "", status="running2")
|
| 777 |
-
|
| 778 |
-
return html, 1, gr.update(value=label)
|
| 779 |
elif next_step == 2:
|
| 780 |
-
html = render_chain(s1, s2, "", status="running3")
|
| 781 |
-
|
| 782 |
-
return html, 2, gr.update(value=label)
|
| 783 |
else:
|
| 784 |
-
html = render_chain(s1, s2, s3, status="done")
|
| 785 |
return html, -1, gr.update(value="βΊ Reset")
|
| 786 |
|
| 787 |
|
|
@@ -809,7 +773,6 @@ def on_select_reset(agent_id):
|
|
| 809 |
return map_html, summary, raw_text, demo_text, chain_html, 0, gr.update(value="βΆ Stage 1: Feature Extraction")
|
| 810 |
|
| 811 |
|
| 812 |
-
|
| 813 |
SHOWCASE_AGENTS = sample_agents[:6]
|
| 814 |
|
| 815 |
|
|
@@ -888,6 +851,7 @@ with gr.Blocks(title="HiCoTraj Demo") as app:
|
|
| 888 |
fn=on_agent_click, inputs=agent_hidden,
|
| 889 |
outputs=[agent_cards, map_out, summary_out, raw_out, chain_out, step_state, run_btn]
|
| 890 |
)
|
|
|
|
| 891 |
def on_load(agent_id):
|
| 892 |
map_html, summary, raw_text, _demo_text, chain_html, step, btn = on_select_reset(agent_id)
|
| 893 |
return map_html, summary, raw_text, chain_html, step, btn
|
|
|
|
| 55 |
if os.path.exists(COT_PATH):
|
| 56 |
with open(COT_PATH, "r") as f:
|
| 57 |
cot_raw = json.load(f)
|
|
|
|
| 58 |
records = cot_raw if isinstance(cot_raw, list) else cot_raw.get("inference_results", [])
|
| 59 |
for result in records:
|
| 60 |
cot_by_agent[int(result["agent_id"])] = result
|
|
|
|
| 69 |
s1 = result.get("step1_response", "")
|
| 70 |
s2 = result.get("step2_response", "")
|
| 71 |
s3 = result.get("step3_response", "")
|
| 72 |
+
p1 = result.get("step1_prompt", "")
|
| 73 |
+
p2 = result.get("step2_prompt", "")
|
| 74 |
+
p3 = result.get("step3_prompt", "")
|
| 75 |
+
return s1, s2, s3, p1, p2, p3
|
| 76 |
|
| 77 |
|
| 78 |
# ββ Mobility text builders ββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
|
|
|
| 87 |
obs_end = agent_sp["end_datetime"].max().strftime("%Y-%m-%d")
|
| 88 |
days = (agent_sp["end_datetime"].max() - agent_sp["start_datetime"].min()).days
|
| 89 |
|
|
|
|
| 90 |
act_counts = agent_sp["act_label"].value_counts().head(3)
|
| 91 |
top_acts = ", ".join(f"{a} ({n})" for a, n in act_counts.items())
|
| 92 |
|
|
|
|
| 93 |
agent_sp2 = agent_sp.copy()
|
| 94 |
agent_sp2["hour"] = agent_sp2["start_datetime"].dt.hour
|
| 95 |
def tod(h):
|
|
|
|
| 144 |
|
| 145 |
# ββ HTML reasoning chain ββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 146 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147 |
CHAIN_CSS = """
|
| 148 |
<style>
|
| 149 |
@import url('https://fonts.googleapis.com/css2?family=DM+Mono:wght@400;500&family=DM+Sans:wght@300;400;500;600&display=swap');
|
|
|
|
| 162 |
overflow: hidden;
|
| 163 |
transition: opacity 0.3s, filter 0.3s;
|
| 164 |
}
|
| 165 |
+
.hct-stage.dim { opacity: 0.28; filter: grayscale(0.6); pointer-events: none; }
|
| 166 |
.hct-stage.active { opacity: 1; }
|
| 167 |
|
| 168 |
/* ββ Stage header strip ββ */
|
|
|
|
| 185 |
font-weight: 600;
|
| 186 |
letter-spacing: 0.04em;
|
| 187 |
text-transform: uppercase;
|
| 188 |
+
flex: 1;
|
| 189 |
}
|
| 190 |
|
| 191 |
/* Stage 1 colors */
|
| 192 |
.hct-s1 { background: #f4f6fb; border: 1.5px solid #d4daf0; }
|
| 193 |
.hct-s1 .hct-head { background: #eaecf7; border-bottom: 1px solid #d4daf0; }
|
| 194 |
+
.hct-s1 .hct-num { background: #dde2f3; color: #3a4a80; }
|
| 195 |
.hct-s1 .hct-title { color: #3a4a80; }
|
| 196 |
|
| 197 |
/* Stage 2 colors */
|
| 198 |
.hct-s2 { background: #fdf8f2; border: 1.5px solid #e8d5b8; }
|
| 199 |
.hct-s2 .hct-head { background: #f7ede0; border-bottom: 1px solid #e8d5b8; }
|
| 200 |
+
.hct-s2 .hct-num { background: #f0dcbf; color: #7a4a10; }
|
| 201 |
.hct-s2 .hct-title { color: #7a4a10; }
|
| 202 |
|
| 203 |
/* Stage 3 colors */
|
| 204 |
.hct-s3 { background: #fff6f5; border: 2px solid #d4453a; }
|
| 205 |
.hct-s3 .hct-head { background: #fce8e7; border-bottom: 1px solid #d4453a; }
|
| 206 |
+
.hct-s3 .hct-num { background: #d4453a; color: #fff; }
|
| 207 |
.hct-s3 .hct-title { color: #b0302a; }
|
| 208 |
|
| 209 |
+
/* ββ Prompt pill ββ */
|
| 210 |
+
.hct-prompt-wrap { padding: 0 14px 8px; }
|
| 211 |
+
.hct-prompt-toggle {
|
| 212 |
+
display: inline-flex; align-items: center; gap: 5px;
|
| 213 |
+
font-family: 'DM Mono', monospace; font-size: 9px;
|
| 214 |
+
letter-spacing: 0.08em; text-transform: uppercase;
|
| 215 |
+
padding: 3px 9px; border-radius: 20px; cursor: pointer;
|
| 216 |
+
border: 1px solid currentColor; opacity: 0.45;
|
| 217 |
+
transition: opacity 0.2s; background: transparent;
|
| 218 |
+
}
|
| 219 |
+
.hct-prompt-toggle:hover { opacity: 0.8; }
|
| 220 |
+
.hct-s1 .hct-prompt-toggle { color: #3a4a80; }
|
| 221 |
+
.hct-s2 .hct-prompt-toggle { color: #7a4a10; }
|
| 222 |
+
.hct-s3 .hct-prompt-toggle { color: #b0302a; }
|
| 223 |
+
.hct-prompt-box {
|
| 224 |
+
display: none;
|
| 225 |
+
margin-top: 6px;
|
| 226 |
+
background: rgba(0,0,0,0.03);
|
| 227 |
+
border-radius: 6px;
|
| 228 |
+
padding: 8px 10px;
|
| 229 |
+
font-family: 'DM Mono', monospace;
|
| 230 |
+
font-size: 10px;
|
| 231 |
+
line-height: 1.65;
|
| 232 |
+
color: #556;
|
| 233 |
+
white-space: pre-wrap;
|
| 234 |
+
word-break: break-word;
|
| 235 |
+
max-height: 150px;
|
| 236 |
+
overflow-y: auto;
|
| 237 |
+
border-left: 2px solid currentColor;
|
| 238 |
+
opacity: 0.7;
|
| 239 |
+
}
|
| 240 |
+
.hct-prompt-box.open { display: block; }
|
| 241 |
+
|
| 242 |
/* ββ Body ββ */
|
| 243 |
.hct-body { padding: 12px 14px; }
|
| 244 |
|
| 245 |
/* ββ Arrow connector ββ */
|
| 246 |
.hct-arrow {
|
| 247 |
+
display: flex; align-items: center; gap: 8px;
|
| 248 |
+
padding: 5px 18px; transition: opacity 0.3s;
|
|
|
|
|
|
|
|
|
|
| 249 |
}
|
| 250 |
+
.hct-arrow-line { flex: 1; height: 1px; background: #d8d4ce; }
|
| 251 |
.hct-arrow-label {
|
| 252 |
+
font-family: 'DM Mono', monospace; font-size: 9px;
|
| 253 |
+
color: #b0a898; letter-spacing: 0.08em; text-transform: uppercase;
|
| 254 |
+
white-space: nowrap; background: white;
|
| 255 |
+
padding: 2px 8px; border: 1px solid #e0dbd4; border-radius: 20px;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 256 |
}
|
| 257 |
|
| 258 |
/* ββ Stage 1: Location table ββ */
|
| 259 |
.hct-loc-table {
|
| 260 |
+
width: 100%; border-collapse: collapse;
|
| 261 |
+
font-size: 11.5px; margin-bottom: 10px;
|
|
|
|
|
|
|
| 262 |
}
|
| 263 |
.hct-loc-table th {
|
| 264 |
+
font-family: 'DM Mono', monospace; font-size: 9px; font-weight: 500;
|
| 265 |
+
letter-spacing: 0.1em; text-transform: uppercase; color: #8090b0;
|
| 266 |
+
border-bottom: 1px solid #d4daf0; padding: 3px 6px 5px; text-align: left;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 267 |
}
|
| 268 |
.hct-loc-table th:not(:first-child) { text-align: right; }
|
| 269 |
.hct-loc-table td {
|
| 270 |
+
padding: 5px 6px; color: #2a3050;
|
| 271 |
+
border-bottom: 1px solid #eaecf5; line-height: 1.3;
|
| 272 |
+
}
|
| 273 |
+
.hct-loc-table td:not(:first-child) {
|
| 274 |
+
text-align: right; font-family: 'DM Mono', monospace;
|
| 275 |
+
font-size: 11px; color: #5060a0;
|
| 276 |
}
|
|
|
|
| 277 |
.hct-loc-table tr:last-child td { border-bottom: none; }
|
| 278 |
+
.hct-loc-name {
|
| 279 |
+
font-weight: 500; max-width: 170px; overflow: hidden;
|
| 280 |
+
text-overflow: ellipsis; white-space: nowrap; display: block;
|
| 281 |
+
}
|
| 282 |
.hct-visit-bar-wrap { display: flex; align-items: center; gap: 6px; justify-content: flex-end; }
|
| 283 |
.hct-visit-bar { height: 4px; border-radius: 2px; background: #6878c8; opacity: 0.55; }
|
| 284 |
|
| 285 |
/* ββ Stage 1: Temporal panel ββ */
|
| 286 |
+
.hct-temporal { display: grid; grid-template-columns: 1fr 1fr; gap: 8px; }
|
| 287 |
+
.hct-temp-block { background: #eef0fa; border-radius: 8px; padding: 8px 10px; }
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 288 |
.hct-temp-label {
|
| 289 |
+
font-family: 'DM Mono', monospace; font-size: 9px; font-weight: 500;
|
| 290 |
+
letter-spacing: 0.1em; text-transform: uppercase; color: #7080b0; margin-bottom: 6px;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 291 |
}
|
| 292 |
.hct-seg-row { display: flex; height: 10px; border-radius: 5px; overflow: hidden; margin-bottom: 5px; }
|
| 293 |
+
.hct-seg { transition: width 0.5s; }
|
| 294 |
+
.seg-morning { background: #fbbf24; }
|
| 295 |
+
.seg-afternoon { background: #f97316; }
|
| 296 |
+
.seg-evening { background: #8b5cf6; }
|
| 297 |
+
.seg-night { background: #1e3a5f; }
|
| 298 |
+
.seg-weekday { background: #6878c8; }
|
| 299 |
+
.seg-weekend { background: #e8c080; }
|
| 300 |
.hct-legend { display: flex; flex-wrap: wrap; gap: 4px 10px; }
|
| 301 |
.hct-leg-item { display: flex; align-items: center; gap: 4px; font-size: 10px; color: #5a6080; }
|
| 302 |
.hct-leg-dot { width: 8px; height: 8px; border-radius: 2px; flex-shrink: 0; }
|
| 303 |
.hct-dist-line {
|
| 304 |
+
margin-top: 8px; font-size: 11px; color: #6070a0;
|
| 305 |
+
font-family: 'DM Mono', monospace; padding: 5px 8px;
|
| 306 |
+
background: #eef0fa; border-radius: 6px;
|
| 307 |
+
display: flex; align-items: center; gap: 6px;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 308 |
}
|
| 309 |
|
| 310 |
+
/* ββ Stage 2: 2x2 grid ββ */
|
| 311 |
+
.hct-dim-grid { display: grid; grid-template-columns: 1fr 1fr; gap: 8px; }
|
|
|
|
|
|
|
|
|
|
|
|
|
| 312 |
.hct-dim-card {
|
| 313 |
+
background: #fff; border: 1px solid #e8d5b8;
|
| 314 |
+
border-radius: 8px; padding: 9px 11px;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 315 |
}
|
| 316 |
+
.hct-dim-head { display: flex; align-items: center; gap: 6px; margin-bottom: 5px; }
|
| 317 |
.hct-dim-icon { font-size: 13px; line-height: 1; }
|
| 318 |
.hct-dim-name {
|
| 319 |
+
font-family: 'DM Mono', monospace; font-size: 9px; font-weight: 500;
|
| 320 |
+
letter-spacing: 0.1em; text-transform: uppercase; color: #a07040;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 321 |
}
|
| 322 |
+
.hct-dim-text { font-size: 11px; color: #3a2a10; line-height: 1.55; }
|
| 323 |
.hct-dim-empty { color: #ccc; font-style: italic; }
|
| 324 |
|
| 325 |
+
/* ββ Stage 3 ββ */
|
| 326 |
+
.hct-pred-row { display: flex; align-items: flex-start; gap: 16px; margin-bottom: 10px; }
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 327 |
.hct-pred-badge {
|
| 328 |
+
background: #d4453a; color: white; border-radius: 8px;
|
| 329 |
+
padding: 8px 14px; text-align: center; flex-shrink: 0;
|
|
|
|
|
|
|
|
|
|
|
|
|
| 330 |
}
|
| 331 |
.hct-pred-val { font-size: 18px; font-weight: 600; line-height: 1.2; white-space: nowrap; }
|
| 332 |
+
.hct-pred-sub {
|
| 333 |
+
font-family: 'DM Mono', monospace; font-size: 9px;
|
| 334 |
+
opacity: 0.8; letter-spacing: 0.08em; text-transform: uppercase; margin-top: 2px;
|
| 335 |
+
}
|
| 336 |
+
.hct-conf-col { flex: 1; padding-top: 4px; }
|
| 337 |
+
.hct-conf-label {
|
| 338 |
+
font-family: 'DM Mono', monospace; font-size: 9px; color: #a04040;
|
| 339 |
+
letter-spacing: 0.08em; text-transform: uppercase; margin-bottom: 4px;
|
| 340 |
+
}
|
| 341 |
.hct-conf-track { height: 6px; background: #f0d0cf; border-radius: 3px; overflow: hidden; margin-bottom: 6px; }
|
| 342 |
.hct-conf-fill { height: 100%; background: linear-gradient(90deg, #e74c3c, #8b0000); border-radius: 3px; }
|
| 343 |
.hct-reasoning {
|
| 344 |
+
font-size: 11.5px; color: #4a2020; line-height: 1.6;
|
| 345 |
+
border-left: 3px solid #e8b0ae; padding-left: 10px;
|
|
|
|
|
|
|
|
|
|
| 346 |
}
|
| 347 |
|
| 348 |
+
/* ββ Idle / loading ββ */
|
| 349 |
.hct-idle { font-size: 12px; color: #b0bac8; padding: 6px 0; font-style: italic; }
|
| 350 |
+
.hct-loading { font-size: 12px; padding: 6px 0; display: flex; align-items: center; gap: 8px; }
|
| 351 |
+
.hct-dot {
|
| 352 |
+
width: 6px; height: 6px; border-radius: 50%; display: inline-block;
|
| 353 |
+
animation: hct-pulse 1.2s ease-in-out infinite;
|
| 354 |
}
|
|
|
|
| 355 |
.hct-dot:nth-child(2) { animation-delay: 0.2s; }
|
| 356 |
.hct-dot:nth-child(3) { animation-delay: 0.4s; }
|
| 357 |
@keyframes hct-pulse {
|
| 358 |
0%,100% { opacity: 0.2; transform: scale(0.8); }
|
| 359 |
+
50% { opacity: 1; transform: scale(1.1); }
|
| 360 |
}
|
| 361 |
.hct-s1 .hct-dot { background: #6878c8; }
|
| 362 |
.hct-s2 .hct-dot { background: #c08040; }
|
| 363 |
.hct-s3 .hct-dot { background: #d4453a; }
|
| 364 |
</style>
|
| 365 |
+
<script>
|
| 366 |
+
function hctTogglePrompt(id) {
|
| 367 |
+
var box = document.getElementById(id);
|
| 368 |
+
var btn = document.getElementById(id + '-btn');
|
| 369 |
+
if (!box) return;
|
| 370 |
+
var open = box.classList.toggle('open');
|
| 371 |
+
btn.textContent = open ? 'β² hide prompt' : 'βΌ show prompt';
|
| 372 |
+
}
|
| 373 |
+
</script>
|
| 374 |
"""
|
| 375 |
|
| 376 |
|
| 377 |
def _loading(msg):
|
| 378 |
+
return (f'<div class="hct-loading">'
|
| 379 |
+
f'<span class="hct-dot"></span><span class="hct-dot"></span><span class="hct-dot"></span>'
|
| 380 |
+
f'<span style="color:#8090a0;font-size:12px">{msg}</span></div>')
|
| 381 |
|
| 382 |
|
| 383 |
+
# ββ Parsers βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 384 |
+
|
| 385 |
def _parse_s1(text):
|
| 386 |
+
locations, dur_map, tod, wk, dist = [], {}, {}, {}, None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 387 |
|
| 388 |
for line in text.splitlines():
|
| 389 |
s = line.strip()
|
| 390 |
+
|
| 391 |
+
# Locations: "- Name: N visits/times/time/times each"
|
| 392 |
+
m = re.match(r'-\s+(.+?):\s+(\d+)\s+(?:visit|time)', s, re.IGNORECASE)
|
| 393 |
if m:
|
| 394 |
locations.append((m.group(1).strip(), int(m.group(2))))
|
| 395 |
+
continue
|
| 396 |
+
|
| 397 |
+
# Duration
|
| 398 |
+
m2 = re.match(r'-?\s*(.+?):\s+(?:Average duration of\s*)?([\d.]+)\s+min(?:utes?)?\s+on average', s, re.IGNORECASE)
|
| 399 |
+
if not m2:
|
| 400 |
+
m2 = re.match(r'-?\s*(.+?):\s+Average duration of ([\d.]+)\s+min', s, re.IGNORECASE)
|
| 401 |
if m2:
|
| 402 |
dur_map[m2.group(1).strip()] = float(m2.group(2))
|
| 403 |
+
|
| 404 |
+
# TOD format A: "65% morning, 23% afternoon, 6% evening, 5% night"
|
| 405 |
+
if not tod:
|
| 406 |
+
mA = re.search(r'(\d+)%\s*morning.*?(\d+)%\s*afternoon.*?(\d+)%\s*evening.*?(\d+)%\s*night', s, re.IGNORECASE)
|
| 407 |
+
if mA:
|
| 408 |
+
tod = {'Morning': int(mA.group(1)), 'Afternoon': int(mA.group(2)),
|
| 409 |
+
'Evening': int(mA.group(3)), 'Night': int(mA.group(4))}
|
| 410 |
+
# TOD format B: "morning: 40%, afternoon: 36%, ..."
|
| 411 |
+
if not tod:
|
| 412 |
+
mB = re.search(r'morning[:\s]+(\d+)%.*?afternoon[:\s]+(\d+)%.*?evening[:\s]+(\d+)%.*?night[:\s]+(\d+)%', s, re.IGNORECASE)
|
| 413 |
+
if mB:
|
| 414 |
+
tod = {'Morning': int(mB.group(1)), 'Afternoon': int(mB.group(2)),
|
| 415 |
+
'Evening': int(mB.group(3)), 'Night': int(mB.group(4))}
|
| 416 |
+
# TOD format C: "Afternoon (43%), morning (27%), ..."
|
| 417 |
if not tod:
|
| 418 |
+
parts = re.findall(r'(morning|afternoon|evening|night)\s*\(?(\d+)%\)?', s, re.IGNORECASE)
|
| 419 |
+
if len(parts) >= 3:
|
| 420 |
+
d = {k.capitalize(): int(v) for k, v in parts}
|
| 421 |
+
if all(k in d for k in ['Morning', 'Afternoon', 'Evening']):
|
| 422 |
+
d.setdefault('Night', 0)
|
| 423 |
+
tod = d
|
| 424 |
+
|
| 425 |
# Weekday/weekend
|
| 426 |
if not wk:
|
| 427 |
m4 = re.search(r'(\d+)%\s*weekday.*?(\d+)%\s*weekend', s, re.IGNORECASE)
|
| 428 |
if m4:
|
| 429 |
wk = {'Weekday': int(m4.group(1)), 'Weekend': int(m4.group(2))}
|
| 430 |
+
|
| 431 |
# Distance
|
| 432 |
if not dist:
|
| 433 |
+
m5 = re.search(r'average distance of approximately ([\d.]+)\s*(?:km|miles?)', s, re.IGNORECASE)
|
| 434 |
if m5:
|
| 435 |
dist = float(m5.group(1))
|
| 436 |
|
| 437 |
+
return [(n, v, dur_map.get(n)) for n, v in locations[:7]], tod, wk, dist
|
|
|
|
| 438 |
|
| 439 |
|
| 440 |
def _parse_s2(text):
|
|
|
|
| 441 |
DIMS = {
|
| 442 |
+
'ROUTINE': ['ROUTINE', 'SCHEDULE'],
|
| 443 |
+
'ECONOMIC': ['ECONOMIC', 'SPENDING'],
|
| 444 |
+
'SOCIAL': ['SOCIAL', 'LIFESTYLE'],
|
| 445 |
+
'STABILITY': ['STABILITY', 'REGULARITY', 'CONSISTENCY', 'URBAN'],
|
|
|
|
| 446 |
}
|
| 447 |
+
sections, current_key, current_lines = {}, None, []
|
|
|
|
|
|
|
| 448 |
|
| 449 |
for line in text.splitlines():
|
| 450 |
s = line.strip()
|
|
|
|
| 451 |
mA = re.match(r'^\d+\.\s+([A-Z][A-Z\s&]+?)(?:\s+ANALYSIS|\s+PATTERNS|\s+INDICATORS|\s+CHARACTERISTICS|\s+STABILITY)?:\s*$', s, re.IGNORECASE)
|
|
|
|
| 452 |
mB = re.match(r'^STEP\s+\d+:\s+([A-Z][A-Z\s&]+?)(?:\s+ANALYSIS|\s+PATTERNS|\s+INDICATORS|\s+CHARACTERISTICS|\s+STABILITY)?\s*$', s, re.IGNORECASE)
|
| 453 |
mm = mA or mB
|
| 454 |
if mm:
|
|
|
|
| 459 |
elif current_key and s:
|
| 460 |
if re.match(r'^\d+\.\d+', s):
|
| 461 |
sub = re.sub(r'^\d+\.\d+[^:]*:\s*', '', s)
|
| 462 |
+
if sub: current_lines.append(sub)
|
|
|
|
| 463 |
elif s.startswith('-'):
|
| 464 |
current_lines.append(s.lstrip('-').strip())
|
| 465 |
elif not re.match(r'^\d+\.', s):
|
|
|
|
| 474 |
if any(kw in k for kw in keywords) and txt:
|
| 475 |
sents = re.split(r'(?<=[.!?])\s+', txt.strip())
|
| 476 |
summary = ' '.join(sents[:2])
|
| 477 |
+
result[dim] = summary[:157] + 'β¦' if len(summary) > 160 else summary
|
|
|
|
|
|
|
| 478 |
break
|
| 479 |
return result
|
| 480 |
|
| 481 |
|
| 482 |
def _parse_s3(text):
|
| 483 |
+
pred, conf, r_lines, in_r = '', 0, [], False
|
|
|
|
|
|
|
| 484 |
for line in text.splitlines():
|
| 485 |
s = line.strip()
|
| 486 |
if s.startswith('INCOME_PREDICTION:'):
|
| 487 |
pred = s.replace('INCOME_PREDICTION:', '').strip()
|
| 488 |
elif s.startswith('INCOME_CONFIDENCE:'):
|
| 489 |
+
try: conf = int(re.search(r'\d+', s).group())
|
| 490 |
+
except: pass
|
|
|
|
|
|
|
| 491 |
elif s.startswith('INCOME_REASONING:'):
|
| 492 |
in_r = True
|
| 493 |
r_lines.append(s.replace('INCOME_REASONING:', '').strip())
|
| 494 |
elif in_r:
|
| 495 |
+
if re.match(r'^2\.', s) or s.startswith('INCOME_'): break
|
| 496 |
+
if s: r_lines.append(s)
|
|
|
|
|
|
|
| 497 |
reasoning = ' '.join(r_lines).strip()
|
| 498 |
sents = re.split(r'(?<=[.!?])\s+', reasoning)
|
| 499 |
reasoning = ' '.join(sents[:3])
|
| 500 |
+
return pred, conf, (reasoning[:277] + 'β¦' if len(reasoning) > 280 else reasoning)
|
| 501 |
+
|
| 502 |
+
|
| 503 |
+
def _extract_prompt_instruction(prompt_text, stage):
|
| 504 |
+
if not prompt_text:
|
| 505 |
+
return ''
|
| 506 |
+
key = f'STEP {stage}:'
|
| 507 |
+
idx = prompt_text.find(key)
|
| 508 |
+
if idx != -1:
|
| 509 |
+
return prompt_text[idx:idx + 600].strip()
|
| 510 |
+
# fallback: first meaningful line
|
| 511 |
+
for line in prompt_text.strip().splitlines():
|
| 512 |
+
if line.strip():
|
| 513 |
+
return line.strip()[:300]
|
| 514 |
+
return ''
|
| 515 |
|
| 516 |
|
| 517 |
+
# ββ Body renderers ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 518 |
+
|
| 519 |
def _s1_body(text, active):
|
| 520 |
if not active:
|
| 521 |
return '<div class="hct-idle">Press βΆ to start</div>'
|
| 522 |
if not text:
|
| 523 |
return _loading('Extracting features')
|
| 524 |
+
|
| 525 |
locs, tod, wk, dist = _parse_s1(text)
|
| 526 |
|
|
|
|
| 527 |
max_v = max((v for _, v, _ in locs), default=1)
|
| 528 |
rows = ''
|
| 529 |
for name, visits, dur in locs:
|
| 530 |
bar_w = int(60 * visits / max_v)
|
| 531 |
dur_str = f'{int(dur)}m' if dur else 'β'
|
| 532 |
+
rows += (f'<tr>'
|
| 533 |
+
f'<td><span class="hct-loc-name" title="{name}">{name}</span></td>'
|
| 534 |
+
f'<td><div class="hct-visit-bar-wrap">'
|
| 535 |
+
f'<div class="hct-visit-bar" style="width:{bar_w}px"></div>{visits}</div></td>'
|
| 536 |
+
f'<td>{dur_str}</td></tr>')
|
| 537 |
+
table = (f'<table class="hct-loc-table">'
|
| 538 |
+
f'<thead><tr><th>Location</th><th>Visits</th><th>Avg Stay</th></tr></thead>'
|
| 539 |
+
f'<tbody>{rows}</tbody></table>') if rows else ''
|
| 540 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 541 |
def seg_bar(data, seg_classes):
|
| 542 |
total = sum(data.values()) or 1
|
| 543 |
segs = ''.join(
|
| 544 |
f'<div class="hct-seg {cls}" style="width:{int(100*v/total)}%"></div>'
|
| 545 |
+
for (label, v), cls in zip(data.items(), seg_classes))
|
|
|
|
| 546 |
legend = ''.join(
|
| 547 |
f'<div class="hct-leg-item"><div class="hct-leg-dot {cls}"></div>{label} {v}%</div>'
|
| 548 |
+
for (label, v), cls in zip(data.items(), seg_classes))
|
|
|
|
| 549 |
return f'<div class="hct-seg-row">{segs}</div><div class="hct-legend">{legend}</div>'
|
| 550 |
|
| 551 |
+
tod_panel = (f'<div class="hct-temp-block"><div class="hct-temp-label">Time of Day</div>'
|
| 552 |
+
f'{seg_bar(tod, ["seg-morning","seg-afternoon","seg-evening","seg-night"])}</div>') if tod else ''
|
| 553 |
+
wk_panel = (f'<div class="hct-temp-block"><div class="hct-temp-label">Weekday / Weekend</div>'
|
| 554 |
+
f'{seg_bar(wk, ["seg-weekday","seg-weekend"])}</div>') if wk else ''
|
| 555 |
+
temporal = f'<div class="hct-temporal">{tod_panel}{wk_panel}</div>' if (tod_panel or wk_panel) else ''
|
| 556 |
+
dist_line = f'<div class="hct-dist-line">π Avg trip distance {dist} mi</div>' if dist else ''
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 557 |
|
| 558 |
return table + temporal + dist_line
|
| 559 |
|
|
|
|
| 564 |
if not text:
|
| 565 |
return _loading('Analyzing behavior')
|
| 566 |
dims = _parse_s2(text)
|
| 567 |
+
DIM_META = [('ROUTINE','π','Schedule'), ('ECONOMIC','π°','Economic'),
|
| 568 |
+
('SOCIAL','π₯','Social'), ('STABILITY','π','Stability')]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 569 |
cards = ''
|
| 570 |
for key, icon, label in DIM_META:
|
| 571 |
txt = dims.get(key, '')
|
| 572 |
+
content = (f'<div class="hct-dim-text">{txt}</div>' if txt
|
| 573 |
+
else '<div class="hct-dim-text hct-dim-empty">β</div>')
|
| 574 |
+
cards += (f'<div class="hct-dim-card">'
|
| 575 |
+
f'<div class="hct-dim-head">'
|
| 576 |
+
f'<span class="hct-dim-icon">{icon}</span>'
|
| 577 |
+
f'<span class="hct-dim-name">{label}</span></div>'
|
| 578 |
+
f'{content}</div>')
|
|
|
|
|
|
|
|
|
|
| 579 |
return f'<div class="hct-dim-grid">{cards}</div>'
|
| 580 |
|
| 581 |
|
|
|
|
| 585 |
if not text:
|
| 586 |
return _loading('Inferring demographics')
|
| 587 |
pred, conf, reasoning = _parse_s3(text)
|
| 588 |
+
return (f'<div class="hct-pred-row">'
|
| 589 |
+
f'<div class="hct-pred-badge">'
|
| 590 |
+
f'<div class="hct-pred-val">{pred or "β"}</div>'
|
| 591 |
+
f'<div class="hct-pred-sub">Income</div></div>'
|
| 592 |
+
f'</div>'
|
| 593 |
+
f'<div class="hct-reasoning">{reasoning}</div>')
|
| 594 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 595 |
|
| 596 |
+
# ββ Main renderer βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 597 |
|
| 598 |
+
def render_chain(s1_text, s2_text, s3_text, status="idle",
|
| 599 |
+
s1_prompt="", s2_prompt="", s3_prompt=""):
|
| 600 |
s1_on = status in ("running1", "running2", "running3", "done")
|
| 601 |
s2_on = status in ("running2", "running3", "done")
|
| 602 |
s3_on = status in ("running3", "done")
|
| 603 |
|
|
|
|
| 604 |
s1_body = _s1_body(s1_text if s1_on else '', s1_on)
|
| 605 |
s2_body = _s2_body(s2_text if s2_on else '', s2_on)
|
| 606 |
s3_body = _s3_body(s3_text if s3_on else '', s3_on)
|
| 607 |
|
| 608 |
+
def prompt_pill(pid, prompt_text, stage_num):
|
| 609 |
+
instr = _extract_prompt_instruction(prompt_text, stage_num)
|
| 610 |
+
if not instr:
|
| 611 |
+
return ''
|
| 612 |
+
safe = instr.replace('&', '&').replace('<', '<').replace('>', '>')
|
| 613 |
+
return (f'<div class="hct-prompt-wrap">'
|
| 614 |
+
f'<span id="{pid}-btn" class="hct-prompt-toggle" '
|
| 615 |
+
f'onclick="hctTogglePrompt(\'{pid}\')">'
|
| 616 |
+
f'βΌ show prompt</span>'
|
| 617 |
+
f'<div id="{pid}" class="hct-prompt-box">{safe}</div>'
|
| 618 |
+
f'</div>')
|
| 619 |
+
|
| 620 |
+
def stage(cls, num, title, body, on, prompt_text, stage_num):
|
| 621 |
dim_cls = 'active' if on else 'dim'
|
| 622 |
+
pill = prompt_pill(f'hct-p-{cls}', prompt_text, stage_num) if on and prompt_text else ''
|
| 623 |
+
return (f'<div class="hct-stage hct-{cls} {dim_cls}">'
|
| 624 |
+
f'<div class="hct-head">'
|
| 625 |
+
f'<span class="hct-num">{num}</span>'
|
| 626 |
+
f'<span class="hct-title">{title}</span>'
|
| 627 |
+
f'</div>'
|
| 628 |
+
f'{pill}'
|
| 629 |
+
f'<div class="hct-body">{body}</div>'
|
| 630 |
+
f'</div>')
|
| 631 |
|
| 632 |
def arrow(label, on):
|
| 633 |
op = '1' if on else '0.2'
|
| 634 |
+
return (f'<div class="hct-arrow" style="opacity:{op}">'
|
| 635 |
+
f'<div class="hct-arrow-line"></div>'
|
| 636 |
+
f'<div class="hct-arrow-label">{label}</div>'
|
| 637 |
+
f'<div class="hct-arrow-line"></div></div>')
|
|
|
|
|
|
|
|
|
|
| 638 |
|
| 639 |
+
html = CHAIN_CSS + '<div class="hct-root">'
|
| 640 |
+
html += stage('s1', 'Stage 01', 'Feature Extraction', s1_body, s1_on, s1_prompt, 1)
|
| 641 |
html += arrow('behavioral abstraction', s2_on)
|
| 642 |
+
html += stage('s2', 'Stage 02', 'Behavioral Analysis', s2_body, s2_on, s2_prompt, 2)
|
| 643 |
html += arrow('demographic inference', s3_on)
|
| 644 |
+
html += stage('s3', 'Stage 03', 'Demographic Inference', s3_body, s3_on, s3_prompt, 3)
|
| 645 |
html += '</div>'
|
| 646 |
return html
|
| 647 |
|
|
|
|
| 735 |
|
| 736 |
|
| 737 |
def run_step(agent_id, step):
|
|
|
|
| 738 |
agent_id = int(agent_id)
|
| 739 |
+
s1, s2, s3, p1, p2, p3 = get_cot(agent_id)
|
| 740 |
next_step = step + 1
|
| 741 |
if next_step == 1:
|
| 742 |
+
html = render_chain(s1, "", "", status="running2", s1_prompt=p1)
|
| 743 |
+
return html, 1, gr.update(value="βΆ Stage 2: Behavioral Analysis")
|
|
|
|
| 744 |
elif next_step == 2:
|
| 745 |
+
html = render_chain(s1, s2, "", status="running3", s1_prompt=p1, s2_prompt=p2)
|
| 746 |
+
return html, 2, gr.update(value="βΆ Stage 3: Demographic Inference")
|
|
|
|
| 747 |
else:
|
| 748 |
+
html = render_chain(s1, s2, s3, status="done", s1_prompt=p1, s2_prompt=p2, s3_prompt=p3)
|
| 749 |
return html, -1, gr.update(value="βΊ Reset")
|
| 750 |
|
| 751 |
|
|
|
|
| 773 |
return map_html, summary, raw_text, demo_text, chain_html, 0, gr.update(value="βΆ Stage 1: Feature Extraction")
|
| 774 |
|
| 775 |
|
|
|
|
| 776 |
SHOWCASE_AGENTS = sample_agents[:6]
|
| 777 |
|
| 778 |
|
|
|
|
| 851 |
fn=on_agent_click, inputs=agent_hidden,
|
| 852 |
outputs=[agent_cards, map_out, summary_out, raw_out, chain_out, step_state, run_btn]
|
| 853 |
)
|
| 854 |
+
|
| 855 |
def on_load(agent_id):
|
| 856 |
map_html, summary, raw_text, _demo_text, chain_html, step, btn = on_select_reset(agent_id)
|
| 857 |
return map_html, summary, raw_text, chain_html, step, btn
|