Press n or j to go to the next uncovered block, b, p or k for the previous block.
| 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 | 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 4x 37x 213x 31x 31x 31x 31x 31x 151x 151x 151x 10800x 2448x 2448x 2448x 10800x 10800x 8352x 16992x 8352x 120x 350x 350x 350x 350x 350x 300x 300x 120x 31x 31x 31x 31x 31x 121x 120x 120x 120x 30x 300x 30x 373x 353x 300x 1104x 1104x 1104x 8100x 8100x 1104x 254x 126x 126x 126x 342x 342x 342x 126x 79x 79x 29x 30x 30x 30x 300x 300x 300x 2700x 2700x 612x 240x 30x 30x 30x 30x 30x 30x 30x 30x 30x 30x 254x 254x 29x 254x 254x 254x 254x 254x 30x 30x 1x 31x 31x 31x 31x 31x 31x 30x 30x 30x 1x 1x 1x 234x 234x 217x 202x 29x 234x 217x 202x 29x 29x 28x 28x 28x 27x 25x 29x 600x 29x 29x 29x 29x 29x 263x 263x 263x 263x 234x 234x 234x 29x 29x 263x 263x 263x 31x 31x 14x 14x 14x 112x 29x 29x 29x 14x 14x 14x 17x 120x 120x 120x 15x 31x 232x 29x 31x 31x 234x 234x 50x 31x 50x 32x 18x 18x 18x 31x 2x 2x 31x 28x 17x 1x 219x 342x 31x 31x 31x 17x 17x 14x 263x 31x 31x 17x 14x 11x 31x 31x 31x 31x 19720x 19720x 19720x 31x 31x 2x 2x 31x 263x 234x 31x 234x 31x 31x 263x 34x 34x 32x 32x 32x 1x 31x 31x 31x 31x 34x 34x 2x 4x 36x 36x | /**
* MCP Tool: sentiment_tracker
*
* Track political group sentiment over a configurable time window using a
* weighted blend of (a) seat-share momentum derived from the current EP API
* MEP composition and (b) observed voting cohesion / defection rates derived
* from EP DOCEO roll-call vote (RCV) XML. The previously informational-only
* `timeframe` parameter now drives a real DOCEO window: votes outside the
* `last_month`/`last_quarter`/`last_year` cutoff are excluded, and the tool
* splits the window into halves and sub-windows to compute trend and
* volatility deterministically.
*
* **Intelligence Perspective:** Real RCV-derived cohesion provides direct
* visibility into intra-group dissent, defection patterns and emerging
* fractures — substantially more reliable than pure seat-share proxies.
*
* **Business Perspective:** Supports policy consultancies, financial
* analysts, and corporate affairs teams who need timely, methodology-sound
* reads on group discipline for lobbying strategy and regulatory risk
* assessment.
*
* **Marketing Perspective:** Appeals to journalists, academic researchers,
* and civic-tech developers as a transparent, time-windowed barometer with
* explicit fallback semantics when DOCEO coverage is sparse.
*
* ISMS Policy: SC-002 (Input Validation), AC-003 (Least Privilege),
* AU-002 (Audit Logging). GDPR Article 5(1)(d) — accuracy principle.
*/
import { z } from 'zod';
import { buildToolResponse, buildErrorResponse } from './shared/responseBuilder.js';
import type { ToolResult } from './shared/types.js';
import { fetchAllCurrentMEPs } from '../utils/mepFetcher.js';
import { doceoClient } from '../clients/ep/doceoClient.js';
import { normalizePoliticalGroup } from '../utils/politicalGroupNormalization.js';
import { withTimeoutAndAbort } from '../utils/timeout.js';
import type { LatestVoteRecord } from '../clients/ep/doceoXmlParser.js';
/**
* Zod input schema for the `sentiment_tracker` MCP tool. Optional
* `groupId` filters the analysis to a single political group; `timeframe`
* selects the DOCEO RCV aggregation window (last 30 / 90 / 365 days).
*/
export const SentimentTrackerSchema = z.object({
groupId: z.string()
.min(1)
.max(50)
.optional()
.describe('Political group ID to track (omit for all groups)'),
timeframe: z.enum(['last_month', 'last_quarter', 'last_year'])
.optional()
.default('last_quarter')
.describe('DOCEO roll-call vote aggregation window: last_month (~30d), last_quarter (~90d), or last_year (~365d). Falls back to seat-share-only proxy with confidenceLevel=LOW when DOCEO coverage is insufficient.'),
});
/**
* Validated parameter type for the `sentiment_tracker` tool, inferred
* from {@link SentimentTrackerSchema}.
*/
export type SentimentTrackerParams = z.infer<typeof SentimentTrackerSchema>;
type Timeframe = NonNullable<SentimentTrackerParams['timeframe']>;
type Trend = 'IMPROVING' | 'STABLE' | 'DECLINING' | 'VOLATILE';
interface GroupSentiment {
groupId: string;
sentimentScore: number;
trend: Trend;
volatility: number;
memberCount: number;
cohesionProxy: number;
}
/** Internal type extending GroupSentiment with unrounded cohesion for aggregation. */
interface InternalGroupSentiment extends GroupSentiment {
_rawCohesion: number;
}
interface SentimentShift {
groupId: string;
fromScore: number;
toScore: number;
magnitude: number;
direction: 'POSITIVE' | 'NEGATIVE';
estimatedCause: string;
}
interface SentimentTrackerResult {
timeframe: string;
groupSentiments: GroupSentiment[];
polarizationIndex: number;
consensusTopics: string[];
divisiveTopics: string[];
sentimentShifts: SentimentShift[];
overallParliamentSentiment: number;
computedAttributes: {
mostPositiveGroup: string;
mostNegativeGroup: string;
highestVolatility: string;
trendingSentiment: 'POSITIVE' | 'NEGATIVE' | 'NEUTRAL';
bimodalityIndex: number;
};
dataAvailable: boolean;
confidenceLevel: 'HIGH' | 'MEDIUM' | 'LOW';
dataFreshness: string;
sourceAttribution: string;
methodology: string;
dataQualityWarnings: string[];
}
const KNOWN_POLITICAL_GROUPS = ['EPP', 'S&D', 'Renew', 'Greens/EFA', 'ECR', 'PfE', 'The Left', 'ESN', 'NI'];
/** Per-timeframe DOCEO fetch and aggregation parameters. */
const TIMEFRAME_CONFIG: Record<Timeframe, {
days: number;
maxWeeks: number;
voteBudget: number;
subWindows: number;
}> = {
last_month: { days: 30, maxWeeks: 2, voteBudget: 100, subWindows: 2 },
last_quarter: { days: 90, maxWeeks: 4, voteBudget: 200, subWindows: 3 },
last_year: { days: 365, maxWeeks: 8, voteBudget: 300, subWindows: 4 },
} as const;
/** Minimum DOCEO RCV votes within the window required to upgrade out of LOW confidence. */
const MIN_RCV_VOTES_FOR_MEDIUM = 10;
const MIN_RCV_VOTES_FOR_HIGH = 40;
/** DOCEO call timeout (mirrors `analyzeCoalitionDynamics` budget). */
const DOCEO_TIMEOUT_MS = 4_000;
/** In-memory cache TTL for the (timeframe-keyed) DOCEO window aggregate. */
const DOCEO_CACHE_TTL_MS = 5 * 60_000;
/** Cohesion bands used to surface DOCEO vote subjects as consensus/divisive topics. */
const CONSENSUS_TOPIC_COHESION_MIN = 0.95;
const DIVISIVE_TOPIC_COHESION_MAX = 0.55;
/** Sentiment score component weights (must sum to 1.0). */
const SCORE_WEIGHTS = {
cohesion: 0.5,
dissent: 0.2,
seatShareMomentum: 0.3,
} as const;
/** Seat-share thresholds used only on the fallback path (no DOCEO data). */
const SEAT_SHARE_THRESHOLDS = { large: 0.25, medium: 0.15, small: 0.05 } as const;
const SEAT_SHARE_FALLBACK_SCORES = { large: 0.3, medium: 0.2, small: 0.1, micro: -0.1 } as const;
/** Trend classification thresholds. */
const TREND_DELTA_THRESHOLD = 0.05;
const VOLATILITY_VARIANCE_THRESHOLD = 0.1;
interface DoceoWindowAggregate {
/** Per-group cohesion stats for the full window. */
groupStats: Map<string, { cohesion: number; rcvVotes: number; topics: string[] }>;
/** Per-group cohesion for the first half of the window (oldest). */
firstHalfCohesion: Map<string, number>;
/** Per-group cohesion for the second half of the window (newest). */
secondHalfCohesion: Map<string, number>;
/** Per-group cohesion variance across N sub-windows (drives VOLATILE classification). */
cohesionVariance: Map<string, number>;
/** Per-group consensus vote subjects (cohesion ≥ CONSENSUS_TOPIC_COHESION_MIN). */
consensusTopics: string[];
/** Per-group divisive vote subjects (cohesion ≤ DIVISIVE_TOPIC_COHESION_MAX). */
divisiveTopics: string[];
/** Total RCV votes inspected within the window. */
rcvVoteCount: number;
/** True when DOCEO returned no RCV-backed records inside the window. */
empty: boolean;
}
let doceoWindowCache:
| { expiresAt: number; key: string; value: DoceoWindowAggregate }
| undefined;
/**
* Test-only hook for resetting the bounded DOCEO window cache between specs.
* @internal
*/
export function clearSentimentTrackerDoceoCache(): void {
doceoWindowCache = undefined;
}
/**
* Format a `Date` as `YYYY-MM-DD` in UTC.
*/
function formatDateIsoUtc(d: Date): string {
return d.toISOString().slice(0, 10);
}
/**
* Compute the Monday of the plenary week containing `now`.
*/
function mostRecentMondayUtc(now: Date): string {
const day = now.getUTCDay();
const daysBack = day === 0 ? 6 : day - 1;
const monday = new Date(now);
monday.setUTCDate(now.getUTCDate() - daysBack);
return formatDateIsoUtc(monday);
}
/**
* Subtract `days` UTC days from a `YYYY-MM-DD` date.
*/
function subtractDaysIso(isoDate: string, days: number): string {
const d = new Date(`${isoDate}T00:00:00Z`);
d.setUTCDate(d.getUTCDate() - days);
return formatDateIsoUtc(d);
}
/**
* Compute majority cohesion for a single group breakdown row.
* Cohesion = `max(for, against, abstain) / total` ∈ [0, 1].
* Returns `null` when no votes were recorded for the group on this vote.
*/
function rowCohesion(
row: { for: number; against: number; abstain: number } | undefined
): number | null {
if (row === undefined) return null;
const total = row.for + row.against + row.abstain;
Iif (total === 0) return null;
return Math.max(row.for, row.against, row.abstain) / total;
}
/**
* Locate the cohesion contribution for `normalizedGroup` in a raw DOCEO group
* breakdown record. Falls back to alias-aware matching via {@link normalizePoliticalGroup}.
*/
function resolveGroupRow(
breakdown: Record<string, { for: number; against: number; abstain: number }> | undefined,
normalizedGroup: string
): { for: number; against: number; abstain: number } | undefined {
Iif (breakdown === undefined) return undefined;
if (normalizedGroup in breakdown) return breakdown[normalizedGroup];
for (const [rawKey, row] of Object.entries(breakdown)) {
Iif (normalizePoliticalGroup(rawKey) === normalizedGroup) return row;
}
return undefined;
}
/**
* Process a single DOCEO response page, filtering RCV records by date and
* appending them to `collected` until the vote budget is reached.
* Returns `{ stop: true }` when the budget is exhausted.
* Records with missing/empty dates are treated as out-of-window and skipped.
* Pre-cutoff records are filtered out but do NOT terminate iteration (DOCEO
* responses are not guaranteed to be sorted newest→oldest within a week).
*/
function ingestDoceoPage(
response: { data: readonly LatestVoteRecord[] },
cutoff: string,
voteBudget: number,
collected: LatestVoteRecord[]
): { stop: boolean } {
for (const vote of response.data) {
Iif (vote.dataSource !== 'RCV') continue;
Iif (vote.groupBreakdown === undefined) continue;
const date = vote.sittingDate ?? vote.date;
// Treat missing/empty dates as out-of-window (skip without stopping).
Iif (date === '') continue;
// Filter out pre-cutoff votes without stopping iteration.
if (date < cutoff) continue;
collected.push(vote);
Iif (collected.length >= voteBudget) return { stop: true };
}
return { stop: false };
}
/**
* Iterate plenary weeks backwards from the current week, fetching DOCEO RCV
* records and filtering them to the time window `[cutoff, now]`. Stops on
* vote-budget or `maxWeeks` boundary, whichever comes first.
*/
async function fetchWindowedDoceoVotes(
timeframe: Timeframe,
abortSignal: AbortSignal,
now: Date = new Date()
): Promise<LatestVoteRecord[]> {
const cfg = TIMEFRAME_CONFIG[timeframe];
const cutoff = subtractDaysIso(formatDateIsoUtc(now), cfg.days);
const collected: LatestVoteRecord[] = [];
let weekStart = mostRecentMondayUtc(now);
for (let weekIndex = 0; weekIndex < cfg.maxWeeks && collected.length < cfg.voteBudget; weekIndex++) {
const response = await doceoClient.getLatestVotes({
weekStart,
includeIndividualVotes: false,
limit: 100,
abortSignal,
});
const { stop } = ingestDoceoPage(response, cutoff, cfg.voteBudget, collected);
Iif (stop) break;
weekStart = subtractDaysIso(weekStart, 7);
}
return collected.slice(0, cfg.voteBudget);
}
/** Sort vote records ascending by sitting date (oldest first), stable on input order ties. */
function sortVotesByDateAscending(votes: LatestVoteRecord[]): LatestVoteRecord[] {
const indexed = votes.map((v, i) => ({ v, i, d: v.sittingDate ?? v.date }));
indexed.sort((a, b) => {
if (a.d === b.d) return a.i - b.i;
return a.d < b.d ? -1 : 1;
});
return indexed.map(x => x.v);
}
/** Compute mean cohesion for a given group over a slice of votes. */
function meanCohesion(
votes: readonly LatestVoteRecord[],
group: string
): { mean: number; count: number } {
let sum = 0;
let count = 0;
for (const vote of votes) {
const c = rowCohesion(resolveGroupRow(vote.groupBreakdown, group));
if (c !== null) { sum += c; count++; }
}
return { mean: count > 0 ? sum / count : 0, count };
}
/** Compute sample variance across non-empty sub-window cohesion means for a group. */
function subWindowVariance(
votes: readonly LatestVoteRecord[],
group: string,
subWindows: number
): number {
if (votes.length < subWindows) return 0;
const bucketSize = Math.ceil(votes.length / subWindows);
const means: number[] = [];
for (let i = 0; i < subWindows; i++) {
const slice = votes.slice(i * bucketSize, (i + 1) * bucketSize);
const { mean, count } = meanCohesion(slice, group);
if (count > 0) means.push(mean);
}
if (means.length < 2) return 0;
const avg = means.reduce((s, v) => s + v, 0) / means.length;
const variance = means.reduce((s, v) => s + (v - avg) ** 2, 0) / means.length;
return variance;
}
/** Identify per-vote consensus / divisive subjects for any analyzed group. */
function extractTopics(
votes: readonly LatestVoteRecord[],
analyzedGroups: readonly string[]
): { consensus: string[]; divisive: string[] } {
const consensus = new Set<string>();
const divisive = new Set<string>();
for (const vote of votes) {
const subject = vote.subject.trim();
Iif (subject === '') continue;
for (const group of analyzedGroups) {
const c = rowCohesion(resolveGroupRow(vote.groupBreakdown, group));
if (c === null) continue;
if (c >= CONSENSUS_TOPIC_COHESION_MIN) consensus.add(subject);
else if (c <= DIVISIVE_TOPIC_COHESION_MAX) divisive.add(subject);
}
}
return {
consensus: [...consensus].slice(0, 10),
divisive: [...divisive].slice(0, 10),
};
}
/**
* Build per-group cohesion statistics, half-window splits and sub-window
* variance from the windowed DOCEO RCV slice.
*/
function buildDoceoAggregate(
votes: LatestVoteRecord[],
analyzedGroups: readonly string[],
subWindows: number
): DoceoWindowAggregate {
const sorted = sortVotesByDateAscending(votes);
const mid = Math.floor(sorted.length / 2);
const firstHalf = sorted.slice(0, mid);
const secondHalf = sorted.slice(mid);
const groupStats = new Map<string, { cohesion: number; rcvVotes: number; topics: string[] }>();
const firstHalfCohesion = new Map<string, number>();
const secondHalfCohesion = new Map<string, number>();
const cohesionVariance = new Map<string, number>();
for (const group of analyzedGroups) {
const fullStats = meanCohesion(sorted, group);
if (fullStats.count > 0) {
groupStats.set(group, { cohesion: fullStats.mean, rcvVotes: fullStats.count, topics: [] });
}
const firstStats = meanCohesion(firstHalf, group);
if (firstStats.count > 0) firstHalfCohesion.set(group, firstStats.mean);
const secondStats = meanCohesion(secondHalf, group);
if (secondStats.count > 0) secondHalfCohesion.set(group, secondStats.mean);
cohesionVariance.set(group, subWindowVariance(sorted, group, subWindows));
}
const { consensus, divisive } = extractTopics(sorted, analyzedGroups);
return {
groupStats,
firstHalfCohesion,
secondHalfCohesion,
cohesionVariance,
consensusTopics: consensus,
divisiveTopics: divisive,
rcvVoteCount: sorted.length,
empty: sorted.length === 0,
};
}
function emptyDoceoAggregate(): DoceoWindowAggregate {
return {
groupStats: new Map(),
firstHalfCohesion: new Map(),
secondHalfCohesion: new Map(),
cohesionVariance: new Map(),
consensusTopics: [],
divisiveTopics: [],
rcvVoteCount: 0,
empty: true,
};
}
/**
* Fetch the DOCEO window aggregate for the requested timeframe with a TTL
* cache. Failures are swallowed — the caller falls back to the seat-share
* proxy path with `confidenceLevel: 'LOW'`.
*/
async function loadDoceoWindowAggregate(
timeframe: Timeframe,
analyzedGroups: readonly string[]
): Promise<DoceoWindowAggregate> {
const now = Date.now();
const key = `${timeframe}::${analyzedGroups.join(',')}`;
Iif (doceoWindowCache?.key === key && doceoWindowCache.expiresAt > now) {
return doceoWindowCache.value;
}
try {
const votes = await withTimeoutAndAbort(
(signal) => fetchWindowedDoceoVotes(timeframe, signal),
DOCEO_TIMEOUT_MS,
'DOCEO window aggregation timed out'
);
const aggregate = buildDoceoAggregate(votes, analyzedGroups, TIMEFRAME_CONFIG[timeframe].subWindows);
doceoWindowCache = { expiresAt: now + DOCEO_CACHE_TTL_MS, key, value: aggregate };
return aggregate;
} catch {
const aggregate = emptyDoceoAggregate();
doceoWindowCache = { expiresAt: now + DOCEO_CACHE_TTL_MS, key, value: aggregate };
return aggregate;
}
}
/** Seat-share fallback score used when DOCEO data is insufficient. */
function fallbackSeatShareScore(memberCount: number, totalMEPs: number): number {
const seatShare = totalMEPs > 0 ? memberCount / totalMEPs : 0;
if (seatShare > SEAT_SHARE_THRESHOLDS.large) return SEAT_SHARE_FALLBACK_SCORES.large;
if (seatShare > SEAT_SHARE_THRESHOLDS.medium) return SEAT_SHARE_FALLBACK_SCORES.medium;
if (seatShare > SEAT_SHARE_THRESHOLDS.small) return SEAT_SHARE_FALLBACK_SCORES.small;
return SEAT_SHARE_FALLBACK_SCORES.micro;
}
/** Seat-share-only trend fallback (no DOCEO half-window data). */
function fallbackSeatShareTrend(seatShare: number): Trend {
if (seatShare > 0.25) return 'STABLE';
if (seatShare > 0.15) return 'IMPROVING';
if (seatShare > 0.05) return 'STABLE';
return 'DECLINING';
}
/**
* Classify a per-group trend using DOCEO half-window cohesion delta and
* sub-window variance. Returns `null` when the DOCEO data is insufficient.
*/
function classifyDoceoTrend(
firstHalf: number | undefined,
secondHalf: number | undefined,
variance: number
): Trend | null {
if (variance > VOLATILITY_VARIANCE_THRESHOLD) return 'VOLATILE';
Iif (firstHalf === undefined || secondHalf === undefined) return null;
const delta = secondHalf - firstHalf;
if (delta > TREND_DELTA_THRESHOLD) return 'IMPROVING';
if (delta < -TREND_DELTA_THRESHOLD) return 'DECLINING';
return 'STABLE';
}
/** Clamp a real-valued score into `[-1, +1]`. */
function clampScore(value: number): number {
return Math.max(-1, Math.min(1, value));
}
/** Round a number to 2 decimal places. */
function round2(value: number): number {
return Math.round(value * 100) / 100;
}
/**
* Compute the per-group sentiment score as a weighted blend of:
* - DOCEO cohesion (centred at 0.5 → score range `[-0.5, +0.5]`),
* - dissent / defection rate (inverse cohesion penalty, range `[-0.5, +0.5]`),
* - seat-share momentum (linear in seat share, range `[-0.5, +0.5]`).
*/
function computeBlendedScore(
cohesion: number,
seatShare: number
): number {
// Cohesion: 0.5 cohesion → 0 contribution, 1.0 → +0.5, 0.0 → -0.5.
const cohesionTerm = (cohesion - 0.5) * 2 * 0.5;
// Dissent: penalty proportional to (1 - cohesion), inverted.
const dissentTerm = (0.5 - (1 - cohesion)) * 0.5;
// Seat-share momentum: 0% → -0.5, 50% → +0.5.
const seatShareTerm = (seatShare - 0.25) * 2;
const raw =
SCORE_WEIGHTS.cohesion * cohesionTerm +
SCORE_WEIGHTS.dissent * dissentTerm +
SCORE_WEIGHTS.seatShareMomentum * seatShareTerm;
return clampScore(raw);
}
interface BuildGroupSentimentArgs {
groupId: string;
memberCount: number;
totalMEPs: number;
doceo: DoceoWindowAggregate;
hasDoceoCoverage: boolean;
}
/** Build the per-group sentiment record using DOCEO data when available. */
function buildGroupSentiment(args: BuildGroupSentimentArgs): InternalGroupSentiment {
const { groupId, memberCount, totalMEPs, doceo, hasDoceoCoverage } = args;
const seatShare = totalMEPs > 0 ? memberCount / totalMEPs : 0;
const stats = doceo.groupStats.get(groupId);
if (!hasDoceoCoverage || stats === undefined) {
const fallbackScore = fallbackSeatShareScore(memberCount, totalMEPs);
const rawCohesion = 0.5 + fallbackScore * 0.3;
return {
groupId,
sentimentScore: round2(fallbackScore),
trend: fallbackSeatShareTrend(seatShare),
volatility: 0.12,
memberCount,
cohesionProxy: round2(rawCohesion),
_rawCohesion: rawCohesion,
};
}
const cohesion = stats.cohesion;
const variance = doceo.cohesionVariance.get(groupId) ?? 0;
const trend = classifyDoceoTrend(
doceo.firstHalfCohesion.get(groupId),
doceo.secondHalfCohesion.get(groupId),
variance
) ?? fallbackSeatShareTrend(seatShare);
const score = computeBlendedScore(cohesion, seatShare);
return {
groupId,
sentimentScore: round2(score),
trend,
volatility: round2(Math.sqrt(variance)),
memberCount,
cohesionProxy: round2(cohesion),
_rawCohesion: cohesion,
};
}
/**
* Compute polarization as `1 − seat-share-weighted mean of cohesion across
* analyzed groups that have observed DOCEO cohesion`. Falls back to
* seat-share-score dispersion when DOCEO cohesion is unavailable. Only groups
* with actual DOCEO observations are weighted into the polarization index;
* groups whose `cohesionProxy` is a fallback value are excluded so the metric
* remains a faithful measure of observed group cohesion.
*/
function computePolarizationIndex(
sentiments: InternalGroupSentiment[],
totalMEPs: number,
hasDoceoCoverage: boolean,
doceo: DoceoWindowAggregate
): number {
Iif (sentiments.length === 0) return 0;
if (hasDoceoCoverage && totalMEPs > 0) {
let weightedCohesion = 0;
let totalWeight = 0;
for (const g of sentiments) {
if (!doceo.groupStats.has(g.groupId)) continue;
const weight = g.memberCount / totalMEPs;
weightedCohesion += g._rawCohesion * weight;
totalWeight += weight;
}
Iif (totalWeight === 0) return 0;
const mean = weightedCohesion / totalWeight;
return round2(Math.max(0, Math.min(1, 1 - mean)));
}
// Fallback: dispersion of fallback seat-share scores.
if (sentiments.length < 2) return 0;
const scores = sentiments.map(s => s.sentimentScore);
const mean = scores.reduce((s, v) => s + v, 0) / scores.length;
const variance = scores.reduce((s, v) => s + (v - mean) ** 2, 0) / scores.length;
return Math.min(1, Math.round(Math.sqrt(variance) * 200) / 100);
}
function computeBimodalityIndex(sentiments: number[]): number {
if (sentiments.length < 3) return 0;
const extreme = sentiments.filter(s => s > 0.1 || s < -0.1).length;
return Math.round((extreme / sentiments.length) * 100) / 100;
}
function buildSentimentShifts(
sentiments: GroupSentiment[],
hasDoceoCoverage: boolean
): SentimentShift[] {
const shifts: SentimentShift[] = [];
for (const g of sentiments) {
const magnitude = Math.abs(g.sentimentScore);
if (magnitude > 0.15) {
shifts.push({
groupId: g.groupId,
fromScore: 0,
toScore: g.sentimentScore,
magnitude: Math.round(magnitude * 100) / 100,
direction: g.sentimentScore > 0 ? 'POSITIVE' : 'NEGATIVE',
estimatedCause: estimateShiftCause(g, hasDoceoCoverage),
});
}
}
return shifts.sort((a, b) => b.magnitude - a.magnitude).slice(0, 5);
}
function estimateShiftCause(g: GroupSentiment, hasDoceoCoverage: boolean): string {
if (!hasDoceoCoverage) {
return g.sentimentScore > 0
? 'Large seat share indicates strong institutional position on tracked topics'
: 'Smaller seat share indicates weaker institutional position or limited leverage on tracked topics';
}
Iif (g.trend === 'VOLATILE') {
return 'High variance in DOCEO group cohesion across sub-windows';
}
Eif (g.sentimentScore > 0) {
return `High DOCEO cohesion (${String(g.cohesionProxy)}) combined with seat-share position`;
}
return `Low DOCEO cohesion (${String(g.cohesionProxy)}) signals dissent / defection within the window`;
}
function classifyTrendingSentiment(score: number): 'POSITIVE' | 'NEGATIVE' | 'NEUTRAL' {
if (score > 0.1) return 'POSITIVE';
Iif (score < -0.1) return 'NEGATIVE';
return 'NEUTRAL';
}
function determineConfidenceLevel(rcvVoteCount: number): 'HIGH' | 'MEDIUM' | 'LOW' {
if (rcvVoteCount >= MIN_RCV_VOTES_FOR_HIGH) return 'HIGH';
if (rcvVoteCount >= MIN_RCV_VOTES_FOR_MEDIUM) return 'MEDIUM';
return 'LOW';
}
function buildEmptySentimentResult(params: SentimentTrackerParams): SentimentTrackerResult {
return {
timeframe: params.timeframe,
groupSentiments: [],
polarizationIndex: 0,
consensusTopics: [],
divisiveTopics: [],
sentimentShifts: [],
overallParliamentSentiment: 0,
computedAttributes: {
mostPositiveGroup: 'N/A',
mostNegativeGroup: 'N/A',
highestVolatility: 'N/A',
trendingSentiment: 'NEUTRAL',
bimodalityIndex: 0
},
dataAvailable: false,
confidenceLevel: 'LOW',
dataFreshness: 'No data available from EP API',
sourceAttribution: 'European Parliament Open Data Portal - data.europarl.europa.eu',
methodology: 'Sentiment derived from seat-share proxies — no MEP data available.',
dataQualityWarnings: ['No MEP data available from EP API — all sentiment scores are empty'],
};
}
function buildSentimentComputedAttrs(
validSentiments: GroupSentiment[],
sentimentScores: number[],
overallSentiment: number
): SentimentTrackerResult['computedAttributes'] {
const sortedBySentiment = [...validSentiments].sort((a, b) => b.sentimentScore - a.sentimentScore);
const sortedByVolatility = [...validSentiments].sort((a, b) => b.volatility - a.volatility);
return {
mostPositiveGroup: sortedBySentiment[0]?.groupId ?? 'N/A',
mostNegativeGroup: sortedBySentiment[sortedBySentiment.length - 1]?.groupId ?? 'N/A',
highestVolatility: sortedByVolatility[0]?.groupId ?? 'N/A',
trendingSentiment: classifyTrendingSentiment(overallSentiment),
bimodalityIndex: computeBimodalityIndex(sentimentScores)
};
}
function buildMethodology(timeframe: Timeframe, hasDoceoCoverage: boolean, rcvVoteCount: number): string {
const window = `${String(TIMEFRAME_CONFIG[timeframe].days)}d`;
if (!hasDoceoCoverage) {
const coverageNote = rcvVoteCount > 0
? `insufficient DOCEO RCV coverage in ${window} window — only ${String(rcvVoteCount)} usable RCVs, minimum ${String(MIN_RCV_VOTES_FOR_MEDIUM)} required`
: `no DOCEO RCV votes found in ${window} window`;
return [
`FALLBACK PATH (${coverageNote}).`,
'Sentiment scores fall back to seat-share proxy (larger group → stronger institutional position).',
'Trend defaults to seat-share band. polarizationIndex = std-dev of fallback scores.',
'Data source (composition): https://data.europarl.europa.eu/api/v2/meps',
].join(' ');
}
return [
`Sentiment score = ${String(SCORE_WEIGHTS.cohesion)} · DOCEO cohesion (centred at 0.5)`,
`+ ${String(SCORE_WEIGHTS.dissent)} · inverse-dissent term`,
`+ ${String(SCORE_WEIGHTS.seatShareMomentum)} · seat-share momentum, clamped to [-1, +1].`,
`Trend derived from half-window cohesion delta (>${String(TREND_DELTA_THRESHOLD)} → IMPROVING/DECLINING) `,
`with VOLATILE when sub-window cohesion variance > ${String(VOLATILITY_VARIANCE_THRESHOLD)}.`,
`polarizationIndex = 1 − seat-share-weighted mean(cohesion) across analyzed groups.`,
`Window: ${window}, ${String(rcvVoteCount)} DOCEO RCVs inspected.`,
'Data sources: https://data.europarl.europa.eu/api/v2/meps (composition); EP DOCEO RCV XML (cohesion).',
].join(' ');
}
interface SentimentBuildContext {
groupCounts: Map<string, number>;
totalMEPs: number;
doceo: DoceoWindowAggregate;
hasDoceoCoverage: boolean;
}
function buildSentimentScores(
targetGroups: readonly string[],
ctx: SentimentBuildContext
): InternalGroupSentiment[] {
return targetGroups.map(gId => buildGroupSentiment({
groupId: gId,
memberCount: ctx.groupCounts.get(gId) ?? 0,
totalMEPs: ctx.totalMEPs,
doceo: ctx.doceo,
hasDoceoCoverage: ctx.hasDoceoCoverage,
}));
}
function buildDataQualityWarnings(
hasDoceoCoverage: boolean,
rcvVoteCount: number,
mepDataComplete: boolean,
mepFailureOffset: number | undefined
): string[] {
const warnings: string[] = [];
if (!hasDoceoCoverage) {
warnings.push(
`DOCEO RCV coverage insufficient (${String(rcvVoteCount)} RCVs in window) — sentiment scores fall back to seat-share proxy`
);
} else if (rcvVoteCount < MIN_RCV_VOTES_FOR_HIGH) {
warnings.push(
`DOCEO RCV coverage limited (${String(rcvVoteCount)} RCVs) — scores reflect a small sample of recent votes`
);
}
Iif (!mepDataComplete) {
warnings.push(
`MEP data is incomplete — pagination failed at offset ${String(mepFailureOffset ?? 0)}; results based on partial data`
);
}
return warnings;
}
function tallyGroupCounts(allMeps: readonly { politicalGroup: string }[]): Map<string, number> {
const groupCounts = new Map<string, number>();
for (const mep of allMeps) {
const canonical = normalizePoliticalGroup(mep.politicalGroup);
Iif (canonical === '' || canonical.toLowerCase() === 'unknown') continue;
groupCounts.set(canonical, (groupCounts.get(canonical) ?? 0) + 1);
}
return groupCounts;
}
function resolveTargetGroups(groupId: string | undefined): readonly string[] {
if (groupId === undefined) return KNOWN_POLITICAL_GROUPS;
const normalized = normalizePoliticalGroup(groupId);
return [normalized === '' ? groupId : normalized];
}
interface SentimentResultArgs {
params: SentimentTrackerParams;
fetchResult: { complete: boolean; failureOffset?: number | undefined };
totalMEPs: number;
groupSentiments: InternalGroupSentiment[];
doceo: DoceoWindowAggregate;
hasDoceoCoverage: boolean;
}
function buildSentimentResult(args: SentimentResultArgs): SentimentTrackerResult {
const { params, fetchResult, totalMEPs, groupSentiments, doceo, hasDoceoCoverage } = args;
const validSentiments = groupSentiments.filter(g => g.memberCount > 0);
const sentimentScores = validSentiments.map(g => g.sentimentScore);
const overallSentiment = sentimentScores.length > 0
? round2(sentimentScores.reduce((s, v) => s + v, 0) / sentimentScores.length)
: 0;
const polarizationIndex = computePolarizationIndex(validSentiments, totalMEPs, hasDoceoCoverage, doceo);
return {
timeframe: params.timeframe,
groupSentiments: groupSentiments.map(({ _rawCohesion: _, ...rest }) => rest),
polarizationIndex,
consensusTopics: hasDoceoCoverage ? doceo.consensusTopics : [],
divisiveTopics: hasDoceoCoverage ? doceo.divisiveTopics : [],
sentimentShifts: buildSentimentShifts(validSentiments, hasDoceoCoverage),
overallParliamentSentiment: overallSentiment,
computedAttributes: buildSentimentComputedAttrs(validSentiments, sentimentScores, overallSentiment),
dataAvailable: validSentiments.length > 0,
confidenceLevel: determineConfidenceLevel(doceo.rcvVoteCount),
dataFreshness: hasDoceoCoverage
? `EP API MEP composition + DOCEO RCV XML in last ${String(TIMEFRAME_CONFIG[params.timeframe].days)}d window`
: 'EP API MEP composition only — DOCEO RCV coverage insufficient for this window',
sourceAttribution: 'European Parliament Open Data Portal - data.europarl.europa.eu',
methodology: buildMethodology(params.timeframe, hasDoceoCoverage, doceo.rcvVoteCount),
dataQualityWarnings: buildDataQualityWarnings(
hasDoceoCoverage,
doceo.rcvVoteCount,
fetchResult.complete,
fetchResult.failureOffset
),
};
}
/**
* Compute political-group sentiment scores over a configurable time window.
*
* Implementation of the MCP `sentiment_tracker` tool. Combines current EP API
* MEP composition with DOCEO roll-call vote (RCV) records inside the requested
* `timeframe` window to derive a per-group sentiment score, trend, volatility,
* polarization index, and consensus/divisive topic lists. Falls back to a
* seat-share-only proxy with `confidenceLevel: 'LOW'` and an explicit
* `methodology` note when DOCEO coverage is insufficient.
*
* @param params - Validated tool parameters (see {@link SentimentTrackerSchema})
* @returns A {@link ToolResult} containing the sentiment report or a structured
* error response on failure.
*/
export async function sentimentTracker(params: SentimentTrackerParams): Promise<ToolResult> {
try {
const fetchResult = await fetchAllCurrentMEPs();
const allMeps = fetchResult.meps;
const totalMEPs = allMeps.length;
if (totalMEPs === 0) {
return buildToolResponse(buildEmptySentimentResult(params));
}
const groupCounts = tallyGroupCounts(allMeps);
const targetGroups = resolveTargetGroups(params.groupId);
const doceo = await loadDoceoWindowAggregate(params.timeframe, targetGroups);
const hasDoceoCoverage = !doceo.empty && doceo.rcvVoteCount >= MIN_RCV_VOTES_FOR_MEDIUM;
const groupSentiments = buildSentimentScores(targetGroups, {
groupCounts, totalMEPs, doceo, hasDoceoCoverage,
});
return buildToolResponse(
buildSentimentResult({ params, fetchResult, totalMEPs, groupSentiments, doceo, hasDoceoCoverage })
);
} catch (error: unknown) {
return buildErrorResponse(
error instanceof Error ? error : new Error(String(error)),
'sentiment_tracker'
);
}
}
/**
* MCP tool metadata for `sentiment_tracker` (name, description, and
* JSON Schema for the tool's input). Consumed by the server's tool
* registry to advertise this tool in `ListTools` responses.
*/
export const sentimentTrackerToolMetadata = {
name: 'sentiment_tracker',
description: 'Track political group sentiment over a configurable time window. Combines current EP API MEP composition with DOCEO roll-call vote (RCV) cohesion / defection data aggregated across last_month (~30d), last_quarter (~90d) or last_year (~365d). Returns per-group sentiment scores (-1 to +1), IMPROVING/STABLE/DECLINING/VOLATILE trends derived from half-window and sub-window DOCEO cohesion deltas, polarization index (1 − seat-share-weighted mean cohesion), and consensus/divisive vote subjects. Falls back to seat-share proxy with confidenceLevel=LOW when DOCEO coverage is insufficient.',
inputSchema: {
type: 'object' as const,
properties: {
groupId: {
type: 'string',
description: 'Political group ID to track (omit for all groups)',
minLength: 1,
maxLength: 50
},
timeframe: {
type: 'string',
enum: ['last_month', 'last_quarter', 'last_year'],
description: 'DOCEO RCV aggregation window: last_month (~30d), last_quarter (~90d), or last_year (~365d). Falls back to seat-share-only proxy with confidenceLevel=LOW when DOCEO coverage is insufficient.',
default: 'last_quarter'
},
}
}
};
/**
* MCP `CallTool` handler entry point for `sentiment_tracker`.
*
* Validates the raw input arguments against {@link SentimentTrackerSchema}
* and delegates execution to {@link sentimentTracker}.
*
* @param args - Raw, untrusted MCP `CallTool` arguments
* @returns The same {@link ToolResult} produced by {@link sentimentTracker}
*/
export async function handleSentimentTracker(args: unknown): Promise<ToolResult> {
const params = SentimentTrackerSchema.parse(args);
return sentimentTracker(params);
}
|