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 | 4x 4x 4x 4x 4x 20x 20x 16x 20x 20x 19x 16x 3x 3x 3x 5x 1x 1x 1x 20x 20x 20x 20x 20x 20x 21x 21x 20x 725x 20x 20x 20x 2x 20x 1x 20x 5x 20x 1x 1x 20x 20x 725x 20x 10x 10x 725x 725x 725x 725x 10x 10x 10x 725x 725x 725x 725x 725x 10x 21x 21x 21x 20x 20x 2x 2x 2x 2x 20x 725x 20x 1x 1x 20x 20x 20x 20x 20x 20x 20x 20x 20x 20x 20x 17x 16x 16x 4x 4x 4x 4x 4x 4x 4x 20x 20x 20x 20x 20x 20x 20x 20x 1x 19x 9x 10x 1x 20x 20x 10x 10x 10x 10x 10x 10x 10x 20x 10x 23x 23x 23x 23x 23x 21x 4x 4x 23x 23x 10x 20x 1x 1x 4x | /**
* MCP Tool: detect_voting_anomalies
*
* Flag unusual voting patterns — party defections, sudden alignment shifts,
* abstention spikes, and cross-party movement signals — using authoritative
* DOCEO roll-call (RCV) records.
*
* **Intelligence Perspective:** Anomaly detection tool identifying deviations
* from each MEP's *own* historical voting baseline within the requested
* period. Surfaces early warning signals for party splits, political
* realignments, and emerging cross-party movements. Consumed by
* `correlate_intelligence` and `early_warning_system`.
*
* **Data source:** EP DOCEO XML (`PV-{term}-{date}-RCV_{lang}.xml`) via the
* shared bounded/cached aggregator. Falls back gracefully to LOW confidence
* with `dataQualityWarnings` when DOCEO is unreachable.
*
* ISMS Policy: SC-002 (Input Validation), AC-003 (Least Privilege),
* AU-002 (Audit Logging). GDPR Article 5(1)(c)/(d) — minimisation & accuracy.
*
* @since 0.8.0
*/
import { DetectVotingAnomaliesSchema } from '../schemas/europeanParliament.js';
import { epClient } from '../clients/europeanParliamentClient.js';
import { doceoClient } from '../clients/ep/doceoClient.js';
import { auditLogger, toErrorMessage } from '../utils/auditLogger.js';
import { normalizePoliticalGroup } from '../utils/politicalGroupNormalization.js';
import { withTimeoutAndAbort } from '../utils/timeout.js';
import {
bucketByWeek,
classifyMepVote,
coverageConfidence,
findCrossPartyAlignmentWindows,
findOutlierWeeks,
findWoWShifts,
DEFAULT_CROSS_PARTY_SHARE,
DEFAULT_WOW_THRESHOLD_PP,
DEFAULT_Z_SCORE_THRESHOLD,
} from '../utils/votingBaseline.js';
import { buildToolResponse } from './shared/responseBuilder.js';
import type { ToolResult } from './shared/types.js';
import type { LatestVoteRecord } from '../clients/ep/doceoXmlParser.js';
import type { ClassifiedVote, WeekBucket } from '../utils/votingBaseline.js';
/** Max DOCEO RCV records aggregated per request (2 pages × 100 per page). */
const DOCEO_RCV_FETCH_LIMIT = 200;
/** Per-page limit for DOCEO getLatestVotes (API enforces ≤ 100). */
const DOCEO_PAGE_LIMIT = 100;
/** DOCEO call timeout, mirrors the shared aggregator guard. */
const DOCEO_TIMEOUT_MS = 2_000;
/** Cap MEPs inspected per group/all-MEPs request to bound runtime. */
const MAX_MEPS_PER_REQUEST = 50;
/** Window (hours) for the NEAR_REALTIME freshness label. */
const NEAR_REALTIME_WINDOW_HOURS = 72;
/**
* Voting anomaly surface, preserves the schema consumed by
* `correlate_intelligence` and adds `evidenceVoteIds` for traceability.
*/
interface VotingAnomaly {
type:
| 'PARTY_DEFECTION'
| 'ABSTENTION_SPIKE'
| 'ALIGNMENT_SHIFT'
| 'CROSS_PARTY_ALIGNMENT_SHIFT';
severity: 'HIGH' | 'MEDIUM' | 'LOW';
mepId: string;
mepName: string;
description: string;
metrics: { expectedValue: number; actualValue: number; deviation: number };
detectedDate: string;
/**
* DOCEO `LatestVoteRecord.id` values supporting the anomaly.
* These are the unique vote identifiers returned by `get_latest_votes`
* (e.g. `"RCV-10-2026-01-15-001"`).
*/
evidenceVoteIds: string[];
}
interface VotingAnomalyAnalysis {
period: { from: string; to: string };
targetScope: string;
anomalies: VotingAnomaly[];
summary: { totalAnomalies: number; highSeverity: number; mediumSeverity: number; lowSeverity: number };
computedAttributes: {
anomalyRate: number;
severityIndex: number;
groupStabilityScore: number;
defectionTrend: string;
riskLevel: string;
};
confidenceLevel: 'HIGH' | 'MEDIUM' | 'LOW';
/**
* Source of voting records consumed. Always `'DOCEO'` for this tool —
* matches the closed set used by sibling OSINT tools (`'EP_API' | 'DOCEO'
* | 'EP_API+DOCEO'`). Use {@link dataAvailable} = `false` to discriminate
* the "DOCEO unreachable / empty period" branch.
*/
dataSource: 'DOCEO';
dataFreshness: string;
sourceAttribution: string;
methodology: string;
dataQualityWarnings: string[];
/**
* Minimum per-MEP RCV votes inspected across the analysed scope (worst-case
* coverage). Drives {@link confidenceLevel} so that group/all-MEP scopes
* are not inflated by a single MEP with high coverage.
*/
rcvVotesInspected: number;
/**
* Maximum per-MEP RCV votes inspected across the analysed scope (best-case
* coverage). Surfaced for observability so callers can see the spread
* between the worst- and best-covered MEP.
*/
rcvVotesInspectedMax: number;
mepsAnalyzed: number;
/**
* `false` when DOCEO was unreachable or returned no records for the
* requested period; absent / `true` otherwise. Replaces the previous
* `dataSource: 'NONE'` sentinel so the `dataSource` field stays a closed
* set across OSINT tools.
*/
dataAvailable?: boolean;
}
/** Trend assessment derived from high-severity counts. */
function classifyDefectionTrend(highCount: number): string {
Iif (highCount > 2) return 'INCREASING';
if (highCount > 0) return 'STABLE';
return 'DECREASING';
}
/** Risk-level classifier derived from high-severity counts. */
function classifyRiskLevel(highCount: number): string {
Iif (highCount > 3) return 'CRITICAL';
if (highCount > 1) return 'ELEVATED';
if (highCount > 0) return 'MODERATE';
return 'LOW';
}
/** Map z-score magnitude to severity (preserves prior HIGH/MEDIUM/LOW semantics). */
function severityFromZ(z: number): 'HIGH' | 'MEDIUM' | 'LOW' {
Iif (z >= 3) return 'HIGH';
Iif (z >= 2) return 'MEDIUM';
return 'LOW';
}
/** Map percentage-point deltas to severity. */
function severityFromDelta(delta: number): 'HIGH' | 'MEDIUM' | 'LOW' {
if (delta >= 40) return 'HIGH';
Iif (delta >= 30) return 'MEDIUM';
return 'LOW';
}
/** Map cross-party alignment share to severity. */
function severityFromShare(share: number): 'HIGH' | 'MEDIUM' | 'LOW' {
Eif (share >= 90) return 'HIGH';
if (share >= 75) return 'MEDIUM';
return 'LOW';
}
/**
* Build the human-readable methodology string emitted alongside each
* response. Reports the *derived* thresholds actually applied for the
* request so callers that override `sensitivityThreshold` see the scaled
* values (not the spec defaults).
*
* @internal
*/
function buildMethodologyDescription(
thresholds: SensitivityThresholds,
sensitivity: number
): string {
const z = Math.round(thresholds.zScore * 100) / 100;
const wow = Math.round(thresholds.wowDelta * 100) / 100;
const crossPct = Math.round(thresholds.crossPartyShare * 100);
const isDefault = sensitivity === 0.3;
const scalingNote = isDefault
? '(spec defaults)'
: `(scaled from spec defaults — z≥${String(DEFAULT_Z_SCORE_THRESHOLD)}, WoW≥`
+ `${String(DEFAULT_WOW_THRESHOLD_PP)}pp, cross-party≥`
+ `${String(Math.round(DEFAULT_CROSS_PARTY_SHARE * 100))}% — `
+ `via sensitivityThreshold=${String(sensitivity)})`;
return 'Per-MEP defection / abstention / cross-party alignment anomaly detection on DOCEO RCV '
+ 'records (up to 200 via 2×100 paged fetching). Per-vote group majority resolved by plurality with alphabetical '
+ 'tie-breaking. Each MEP vote classified as aligned / defected / abstained / absent against their '
+ `home-group majority. Anomalies emitted when defection-rate z-score ≥ ${String(z)}, `
+ `abstention-rate z-score ≥ ${String(z)}, week-over-week defection delta ≥ ${String(wow)}pp, `
+ `or non-home-group alignment share ≥ ${String(crossPct)}% in a weekly sub-window `
+ `${scalingNote}. Confidence reflects worst-case RCV coverage across analysed MEPs: `
+ 'HIGH ≥50 votes inspected, MEDIUM 10–49, LOW <10. Source: EP DOCEO XML.';
}
interface MepRef {
id: string;
name: string;
politicalGroup: string;
}
/**
* Per-request thresholds derived from `sensitivityThreshold`.
*
* Defaults match the issue specification: z ≥ 1.5, WoW ≥ 20pp, cross-party
* share ≥ 0.6. The sensitivity parameter scales these so lower values yield
* more anomalies (backward-compatible with prior callers).
*/
interface SensitivityThresholds {
zScore: number;
wowDelta: number;
crossPartyShare: number;
}
function deriveThresholds(sensitivity: number): SensitivityThresholds {
// Linear scale anchored at the default 0.3 → spec defaults.
const factor = sensitivity / 0.3;
return {
zScore: Math.max(0.5, DEFAULT_Z_SCORE_THRESHOLD * factor),
wowDelta: Math.max(5, DEFAULT_WOW_THRESHOLD_PP * factor),
crossPartyShare: Math.min(0.95, Math.max(0.3, DEFAULT_CROSS_PARTY_SHARE * factor)),
};
}
/**
* Detect anomalies for a single MEP from the supplied DOCEO RCV records.
*
* @param mep - MEP reference (id/name/politicalGroup).
* @param votes - All DOCEO RCV records in the inspection window.
* @param thresholds - Sensitivity-adjusted detection thresholds.
* @param detectedDate - Period end date stamped onto anomaly records.
* @returns Detected anomalies and the per-MEP RCV inspection count.
*/
function detectMepAnomaliesFromDoceo(
mep: MepRef,
votes: LatestVoteRecord[],
thresholds: SensitivityThresholds,
detectedDate: string
): { anomalies: VotingAnomaly[]; classified: ClassifiedVote[]; buckets: WeekBucket[] } {
const homeGroup = mep.politicalGroup !== '' ? normalizePoliticalGroup(mep.politicalGroup) : null;
const classified = votes.map(v => classifyMepVote(v, mep.id, homeGroup));
const buckets = bucketByWeek(classified);
const anomalies: VotingAnomaly[] = [];
for (const outlier of findOutlierWeeks(buckets, 'defectionRate', thresholds.zScore)) {
anomalies.push({
type: 'PARTY_DEFECTION',
severity: severityFromZ(outlier.z),
mepId: mep.id,
mepName: mep.name,
description: `Defection rate of ${String(outlier.value)}% during week of ${outlier.weekStart} `
+ `(z=${String(outlier.z)} vs ${mep.politicalGroup || 'home group'} majority on DOCEO RCVs)`,
metrics: {
expectedValue: outlier.baselineMean,
actualValue: outlier.value,
deviation: Math.round((outlier.value - outlier.baselineMean) * 100) / 100,
},
detectedDate,
evidenceVoteIds: outlier.voteIds,
});
}
for (const outlier of findOutlierWeeks(buckets, 'abstentionRate', thresholds.zScore)) {
anomalies.push({
type: 'ABSTENTION_SPIKE',
severity: severityFromZ(outlier.z),
mepId: mep.id,
mepName: mep.name,
description: `Abstention rate of ${String(outlier.value)}% during week of ${outlier.weekStart} `
+ `(z=${String(outlier.z)} vs MEP's own baseline)`,
metrics: {
expectedValue: outlier.baselineMean,
actualValue: outlier.value,
deviation: Math.round((outlier.value - outlier.baselineMean) * 100) / 100,
},
detectedDate,
evidenceVoteIds: outlier.voteIds,
});
}
for (const shift of findWoWShifts(buckets, thresholds.wowDelta)) {
anomalies.push({
type: 'ALIGNMENT_SHIFT',
severity: severityFromDelta(shift.delta),
mepId: mep.id,
mepName: mep.name,
description: `Week-over-week defection rate jumped by ${String(shift.delta)}pp `
+ `(${shift.fromWeek} → ${shift.toWeek})`,
metrics: {
expectedValue: shift.previousRate,
actualValue: Math.round((shift.previousRate + shift.delta) * 100) / 100,
deviation: shift.delta,
},
detectedDate,
evidenceVoteIds: shift.voteIds,
});
}
for (const window of findCrossPartyAlignmentWindows(buckets, thresholds.crossPartyShare)) {
const thresholdPct = Math.round(thresholds.crossPartyShare * 100);
anomalies.push({
type: 'CROSS_PARTY_ALIGNMENT_SHIFT',
severity: severityFromShare(window.sharePercent),
mepId: mep.id,
mepName: mep.name,
description: `Voted with non-home group majorities on ${String(window.sharePercent)}% of `
+ `${String(window.decisive)} decisive RCVs during week of ${window.weekStart}`,
metrics: {
expectedValue: thresholdPct,
actualValue: window.sharePercent,
deviation: Math.round((window.sharePercent - thresholdPct) * 100) / 100,
},
detectedDate,
evidenceVoteIds: window.voteIds,
});
}
return { anomalies, classified, buckets };
}
/** Count RCV votes where the MEP actually appeared on the roll. */
function countVotesInspected(classified: ClassifiedVote[]): number {
let n = 0;
for (const c of classified) Eif (c.alignment !== 'absent') n += 1;
return n;
}
/**
* Determine whether the DOCEO source can be labelled NEAR_REALTIME.
*
* NEAR_REALTIME requires at least one DOCEO record with a sitting date within
* the configured window (default 72h) of `referenceDate`.
*/
function isNearRealtime(votes: LatestVoteRecord[], referenceDate: Date): boolean {
const cutoff = referenceDate.getTime() - NEAR_REALTIME_WINDOW_HOURS * 3_600_000;
for (const v of votes) {
const date = v.sittingDate ?? v.date;
Iif (date === '') continue;
const ts = Date.parse(date);
Iif (!Number.isNaN(ts) && ts >= cutoff) return true;
}
return false;
}
/**
* Return the most recent valid sitting date in the supplied record list,
* or `null` when none of the records have a parseable date.
*
* @internal
*/
function mostRecentSittingDate(votes: LatestVoteRecord[]): Date | null {
let latest: number | null = null;
for (const v of votes) {
const date = v.sittingDate ?? v.date;
Iif (date === '') continue;
const ts = Date.parse(date);
Iif (Number.isNaN(ts)) continue;
if (latest === null || ts > latest) latest = ts;
}
return latest === null ? null : new Date(latest);
}
/**
* Fetch DOCEO RCV records covering the request window. Implements paged
* fetching (2 × 100) since the DOCEO client validates `limit ≤ 100`.
* Returns an empty sentinel when DOCEO is unreachable so the caller can
* surface a graceful `dataQualityWarning` instead of throwing.
*/
async function fetchDoceoRcvRecords(
period: { from: string; to: string }
): Promise<{ records: LatestVoteRecord[]; doceoAvailable: boolean }> {
try {
// Normalize period.to to an ISO date string for DOCEO weekStart.
// DOCEO `weekStart` selects the plenary week containing the supplied date;
// we use the period's *end* date to capture the most recent week in scope.
const firstPage = await withTimeoutAndAbort(
(signal) => doceoClient.getLatestVotes({
includeIndividualVotes: true,
limit: DOCEO_PAGE_LIMIT,
offset: 0,
weekStart: period.to,
abortSignal: signal,
}),
DOCEO_TIMEOUT_MS,
'DOCEO RCV fetch (page 1) timed out'
);
let allRecords = firstPage.data;
// Second page if the first was full — up to DOCEO_RCV_FETCH_LIMIT total.
if (firstPage.data.length >= DOCEO_PAGE_LIMIT) {
try {
const secondPage = await withTimeoutAndAbort(
(signal) => doceoClient.getLatestVotes({
includeIndividualVotes: true,
limit: DOCEO_PAGE_LIMIT,
offset: DOCEO_PAGE_LIMIT,
weekStart: period.to,
abortSignal: signal,
}),
DOCEO_TIMEOUT_MS,
'DOCEO RCV fetch (page 2) timed out'
);
allRecords = [...allRecords, ...secondPage.data];
} catch {
// Second page failure is non-fatal; continue with first page only.
}
}
const records = allRecords
.filter(v => v.dataSource === 'RCV')
.slice(0, DOCEO_RCV_FETCH_LIMIT);
return { records, doceoAvailable: true };
} catch (error: unknown) {
auditLogger.logError('detect_voting_anomalies.doceo_fetch', { from: period.from, to: period.to }, toErrorMessage(error));
return { records: [], doceoAvailable: false };
}
}
interface DetectionResult {
scope: string;
anomalies: VotingAnomaly[];
/** Minimum per-MEP RCV votes inspected (worst-case coverage). */
rcvVotesInspected: number;
/** Maximum per-MEP RCV votes inspected (best-case coverage). */
rcvVotesInspectedMax: number;
mepsAnalyzed: number;
}
/**
* Aggregate per-MEP anomaly detection over a list of MEPs.
*
* Returns *minimum* and *maximum* RCV votes inspected across the MEP list.
* The minimum drives confidence/warnings so that group/all-MEP scopes are not
* inflated by a single high-coverage MEP — when most MEPs have few or zero
* inspected votes (e.g. broad scope, narrow DOCEO window) the response
* correctly reports LOW/MEDIUM confidence instead of HIGH.
*/
function runDetectionForMeps(
meps: MepRef[],
records: LatestVoteRecord[],
thresholds: SensitivityThresholds,
detectedDate: string
): { anomalies: VotingAnomaly[]; rcvVotesInspected: number; rcvVotesInspectedMax: number } {
const allAnomalies: VotingAnomaly[] = [];
const perMepInspected: number[] = [];
for (const mep of meps) {
const { anomalies, classified } = detectMepAnomaliesFromDoceo(mep, records, thresholds, detectedDate);
const inspected = countVotesInspected(classified);
perMepInspected.push(inspected);
allAnomalies.push(...anomalies);
auditLogger.logDataAccess(
'detect_voting_anomalies.mep_analysis',
{ mepId: mep.id, rcvVotesInspected: inspected },
inspected
);
}
// Worst-case coverage drives confidence; best-case is reported separately
// for observability. When no MEPs were analysed both values are 0.
const rcvVotesInspected = perMepInspected.length > 0 ? Math.min(...perMepInspected) : 0;
const rcvVotesInspectedMax = perMepInspected.length > 0 ? Math.max(...perMepInspected) : 0;
return { anomalies: allAnomalies, rcvVotesInspected, rcvVotesInspectedMax };
}
/**
* Detect anomalies for a single MEP.
*
* @internal Exported for unit testing.
*/
async function detectSingleMepAnomalies(
mepId: string,
thresholds: SensitivityThresholds,
period: { from: string; to: string },
records: LatestVoteRecord[]
): Promise<DetectionResult> {
const mep = await epClient.getMEPDetails(mepId);
const { anomalies, rcvVotesInspected, rcvVotesInspectedMax } = runDetectionForMeps(
[{ id: mep.id, name: mep.name, politicalGroup: mep.politicalGroup }],
records,
thresholds,
period.to
);
return { scope: `MEP: ${mepId}`, anomalies, rcvVotesInspected, rcvVotesInspectedMax, mepsAnalyzed: 1 };
}
/** Detect anomalies for a political group or all MEPs (bounded scan). */
async function detectGroupAnomalies(
groupId: string | undefined,
thresholds: SensitivityThresholds,
period: { from: string; to: string },
records: LatestVoteRecord[]
): Promise<DetectionResult> {
const groupFilter: { group?: string } = {};
if (groupId !== undefined) groupFilter.group = groupId;
const scope = groupId !== undefined ? `Group: ${groupId}` : 'All MEPs';
const mepsResult = await epClient.getCurrentMEPs({ ...groupFilter, limit: MAX_MEPS_PER_REQUEST });
const meps: MepRef[] = mepsResult.data.map(m => ({
id: m.id,
name: m.name,
politicalGroup: m.politicalGroup,
}));
const { anomalies, rcvVotesInspected, rcvVotesInspectedMax } =
runDetectionForMeps(meps, records, thresholds, period.to);
return { scope, anomalies, rcvVotesInspected, rcvVotesInspectedMax, mepsAnalyzed: meps.length };
}
/** Roll up anomalies into severity counts and derived attributes. */
function buildAnomalySummary(anomalies: VotingAnomaly[]): {
highSeverity: number;
mediumSeverity: number;
lowSeverity: number;
anomalyRate: number;
severityIndex: number;
} {
const highSeverity = anomalies.filter(a => a.severity === 'HIGH').length;
const mediumSeverity = anomalies.filter(a => a.severity === 'MEDIUM').length;
const lowSeverity = anomalies.filter(a => a.severity === 'LOW').length;
const anomalyRate = anomalies.length > 0
? Math.round((highSeverity / anomalies.length) * 100) / 100
: 0;
const severityIndex = anomalies.length > 0
? Math.round(((highSeverity * 3 + mediumSeverity * 2 + lowSeverity) / anomalies.length) * 100) / 100
: 0;
return { highSeverity, mediumSeverity, lowSeverity, anomalyRate, severityIndex };
}
/**
* Compute data-quality warnings for a given fetch outcome.
*
* @internal
*/
function collectDataQualityWarnings(
doceoAvailable: boolean,
recordCount: number,
rcvVotesInspected: number
): string[] {
const warnings: string[] = [];
if (!doceoAvailable) {
warnings.push('DOCEO RCV source unavailable — anomaly detection deferred (LOW confidence).');
} else if (recordCount === 0) {
warnings.push('No DOCEO RCV records returned for the requested period.');
} else if (rcvVotesInspected < 10) {
warnings.push('Fewer than 10 RCV votes inspected — confidence reduced to LOW.');
}
return warnings;
}
/**
* Resolve the data freshness label.
*
* - `NEAR_REALTIME` when at least one DOCEO sitting falls within
* {@link NEAR_REALTIME_WINDOW_HOURS} of "now".
* - Otherwise, surface the most recent sitting date and the actual age in
* hours so consumers can reason about staleness without inflating the
* window.
*/
function resolveDataFreshness(
dataAvailable: boolean,
records: LatestVoteRecord[]
): string {
if (!dataAvailable) return 'No DOCEO RCV data available for the requested period';
const now = new Date();
Iif (isNearRealtime(records, now)) return 'NEAR_REALTIME';
const latest = mostRecentSittingDate(records);
Iif (latest === null) return `DOCEO RCV data — outside ${String(NEAR_REALTIME_WINDOW_HOURS)}h NEAR_REALTIME window`;
const ageMs = now.getTime() - latest.getTime();
const ageHours = Math.max(0, Math.round(ageMs / 3_600_000));
return `DOCEO RCV data — latest sitting ${String(ageHours)}h old (outside ${String(NEAR_REALTIME_WINDOW_HOURS)}h NEAR_REALTIME window)`;
}
/** Resolve the confidence level from the fetch outcome and coverage. */
function resolveConfidenceLevel(
dataAvailable: boolean,
rcvVotesInspected: number
): 'HIGH' | 'MEDIUM' | 'LOW' {
if (!dataAvailable) return 'LOW';
return coverageConfidence(rcvVotesInspected);
}
/**
* Handles the `detect_voting_anomalies` MCP tool request.
*
* See file-level JSDoc for methodology details. Each anomaly carries
* `evidenceVoteIds` referencing the contributing DOCEO RCV records.
*
* @param args - Raw tool arguments, validated against {@link DetectVotingAnomaliesSchema}
* @returns MCP tool result containing detected anomalies with evidence vote IDs
* @throws If `args` fails schema validation or the EP API is unreachable.
*
* @security Input validated with Zod. Audit logs record only `mepId` and
* sanitised counts (no MEP names or vote contents).
* @since 0.8.0
*/
export async function handleDetectVotingAnomalies(
args: unknown
): Promise<ToolResult> {
const params = DetectVotingAnomaliesSchema.parse(args);
try {
const period = { from: params.dateFrom ?? '2024-01-01', to: params.dateTo ?? '2024-12-31' };
const thresholds = deriveThresholds(params.sensitivityThreshold);
const { records, doceoAvailable } = await fetchDoceoRcvRecords(period);
const result = params.mepId !== undefined
? await detectSingleMepAnomalies(params.mepId, thresholds, period, records)
: await detectGroupAnomalies(params.groupId, thresholds, period, records);
const summary = buildAnomalySummary(result.anomalies);
const dataAvailable = doceoAvailable && records.length > 0;
const analysis: VotingAnomalyAnalysis = {
period,
targetScope: result.scope,
anomalies: result.anomalies,
summary: {
totalAnomalies: result.anomalies.length,
highSeverity: summary.highSeverity,
mediumSeverity: summary.mediumSeverity,
lowSeverity: summary.lowSeverity,
},
computedAttributes: {
anomalyRate: summary.anomalyRate,
severityIndex: summary.severityIndex,
groupStabilityScore: Math.round((1 - summary.severityIndex / 3) * 100 * 100) / 100,
defectionTrend: classifyDefectionTrend(summary.highSeverity),
riskLevel: classifyRiskLevel(summary.highSeverity),
},
confidenceLevel: resolveConfidenceLevel(dataAvailable, result.rcvVotesInspected),
dataSource: 'DOCEO',
dataFreshness: resolveDataFreshness(dataAvailable, records),
sourceAttribution: 'European Parliament DOCEO XML — europarl.europa.eu/doceo',
methodology: buildMethodologyDescription(thresholds, params.sensitivityThreshold),
dataQualityWarnings: collectDataQualityWarnings(doceoAvailable, records.length, result.rcvVotesInspected),
rcvVotesInspected: result.rcvVotesInspected,
rcvVotesInspectedMax: result.rcvVotesInspectedMax,
mepsAnalyzed: result.mepsAnalyzed,
};
if (!dataAvailable) {
analysis.dataAvailable = false;
}
return buildToolResponse(analysis);
} catch (error: unknown) {
const errorMessage = error instanceof Error ? error.message : 'Unknown error';
throw new Error(`Failed to detect voting anomalies: ${errorMessage}`);
}
}
/**
* Tool metadata for MCP registration
*/
export const detectVotingAnomaliesToolMetadata = {
name: 'detect_voting_anomalies',
description: 'Detect unusual voting patterns including party defections, abstention spikes, week-over-week alignment shifts, and cross-party movement signals. Uses DOCEO RCV roll-call records and per-MEP rolling baselines (defection z ≥1.5, abstention z ≥1.5, WoW Δ ≥20pp, cross-party share ≥60%). Returns anomalies with evidenceVoteIds, severity classification (HIGH/MEDIUM/LOW), group stability score, defection trend, and risk level.',
inputSchema: {
type: 'object' as const,
properties: {
mepId: {
type: 'string',
description: 'MEP identifier (omit for broad analysis)',
minLength: 1,
maxLength: 100
},
groupId: {
type: 'string',
description: 'Political group to analyze',
minLength: 1,
maxLength: 50
},
dateFrom: {
type: 'string',
description: 'Analysis start date (YYYY-MM-DD format)',
pattern: '^\\d{4}-\\d{2}-\\d{2}$'
},
dateTo: {
type: 'string',
description: 'Analysis end date (YYYY-MM-DD format)',
pattern: '^\\d{4}-\\d{2}-\\d{2}$'
},
sensitivityThreshold: {
type: 'number',
description: 'Anomaly sensitivity (0-1, lower = more anomalies detected). Default 0.3 matches the spec thresholds (z≥1.5, WoW≥20pp, cross-party≥60%).',
minimum: 0,
maximum: 1,
default: 0.3
}
}
}
};
|