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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 | 4x 4x 4x 230x 230x 230x 226x 226x 226x 226x 226x 115x 17x 17x 115x 9x 106x 70x 17x 28x 28x 28x 36x 19x 9x 9x 115x 115x 115x 46x 46x 46x 69x 115x 115x 62x 9x 9x 53x 115x 115x 115x 115x 115x 115x 115x 115x 115x 115x 115x 115x 118x 115x 69x 10x 10x 115x 115x 115x 115x 115x 115x 115x 115x 115x 38x 34x 38x 118x 115x 115x 115x 40x 37x 3x 83x 9x 9x 83x 9x 74x 36x 36x 83x 83x 28x 83x 83x 83x 28x 36x 28x 9x 9x 83x 83x 83x 83x 36x 36x 36x 36x 36x 36x 83x 36x 39x 36x 36x 36x 36x 5x 36x 25x 36x 18x 36x 36x 36x 36x 36x 36x 36x 74x 37x 37x 37x 37x 37x 37x 37x 37x 37x 13x 37x 37x 36x 126x 126x 126x 8x 118x 36x 36x 36x 36x 115x 115x 69x 115x 115x 115x 115x 36x 37x 37x 37x 37x 37x 37x 36x 36x 36x 36x 37x 37x 115x 83x 37x 37x 83x 83x 37x 37x 38x 38x 38x 36x 1x 1x 4x | /**
* MCP Tool: monitor_legislative_pipeline
*
* Real-time legislative pipeline status with bottleneck detection and
* timeline forecasting backed by the authoritative procedure-event
* lifecycle (`/procedures/{id}/events`).
*
* **Intelligence Perspective:** Pipeline monitoring tool providing situational
* awareness of legislative progress — enables early warning for stalled
* procedures, bottleneck identification, and timeline forecasting based on
* historical median dwell statistics rather than heuristics.
*
* **Methodology:**
* 1. Load the corpus-wide lifecycle model from cache only
* (`getCachedLifecycleStatistics`). The corpus rebuild
* (`/procedures` + up to 500 `/procedures/{id}/events`) is far too
* expensive for the request path under the EP API's 100 req/min rate
* limit — even with a cooperative deadline, in-flight requests cannot be
* cancelled and the budget's worth of token-bucket spend would starve
* the request's own `/procedures` + `/events` fan-out. On a cache miss
* the empty model is used and per-procedure forecasts gracefully degrade
* to `INSUFFICIENT_DATA`. Out-of-band warmup (e.g. background jobs or
* sibling tools whose primary purpose is the corpus itself) populates
* the 30-min cache via `getLifecycleStatistics`.
* 2. For each procedure in scope, fetch its event timeline with bounded
* concurrency (≤8 parallel) and `Promise.allSettled`.
* 3. `daysInCurrentStage` = days between the latest event and `now`.
* 4. `bottleneckRisk` = percentile bucket of the current dwell vs. the
* historical distribution for the same `(procedureType, stage)`.
* 5. `estimatedCompletionDays` = historical median remaining days from the
* current stage (falls back to a heuristic when the corpus lacks data,
* in which case `forecastBasis` is `INSUFFICIENT_DATA`).
*
* ISMS Policy: SC-002 (Input Validation), AC-003 (Least Privilege)
*/
import { MonitorLegislativePipelineSchema } from '../schemas/europeanParliament.js';
import { epClient } from '../clients/europeanParliamentClient.js';
import type { Procedure, EPEvent } from '../types/europeanParliament.js';
import type { ToolResult } from './shared/types.js';
import { withTimeout, TimeoutError } from '../utils/timeout.js';
// (buildTimeoutResponse no longer used directly — see buildPipelineTimeoutResponse)
import { ToolError } from './shared/errors.js';
import {
getCachedLifecycleStatistics,
emptyLifecycleStatisticsModel,
fetchEventsBounded,
sortEventsChronologically,
normalizeStageKey,
lookupStageStatistics,
type LifecycleStatisticsModel,
type StageDwellStatistics,
} from '../utils/lifecycleStatistics.js';
/**
* Maximum wall-clock time (ms) allowed for the full
* `monitor_legislative_pipeline` operation. Chosen to align with sibling
* intelligence tools and stay below the integration-test 120 s ceiling.
*/
const OPERATION_TIMEOUT_MS = 30_000;
/** Forecast basis discriminator emitted in the response envelope. */
export type ForecastBasis = 'HISTORICAL_MEDIAN' | 'INSUFFICIENT_DATA' | 'NOT_APPLICABLE';
/** A single lifecycle event echoed for traceability. */
export interface PipelineLifecycleEvent {
date: string;
stage: string;
rawType: string;
title: string;
}
/** Computed attributes for a single pipeline item */
interface PipelineItemComputedAttrs {
progressPercentage: number;
velocityScore: number;
complexityIndicator: string;
estimatedCompletionDays: number;
bottleneckRisk: string;
}
/** A single procedure in the pipeline */
interface PipelineItem {
procedureId: string;
title: string;
type: string;
currentStage: string;
committee: string;
daysInCurrentStage: number;
isStalled: boolean;
nextExpectedAction: string;
computedAttributes: PipelineItemComputedAttrs;
lifecycleEvents: PipelineLifecycleEvent[];
}
/** Bottleneck analysis */
interface BottleneckInfo {
stage: string;
procedureCount: number;
avgDaysStuck: number;
severity: string;
/** Historical 95th-percentile dwell (days) for the stage — bottleneck threshold. */
thresholdDays: number;
}
/** Full pipeline analysis result */
interface LegislativePipelineAnalysis {
period: { from: string; to: string };
filter: { committee?: string; status: string };
pipeline: PipelineItem[];
summary: {
totalProcedures: number;
activeCount: number;
stalledCount: number;
completedCount: number;
avgDaysInPipeline: number;
};
bottlenecks: BottleneckInfo[];
computedAttributes: {
pipelineHealthScore: number;
throughputRate: number;
bottleneckIndex: number;
stalledProcedureRate: number;
estimatedClearanceTime: number;
legislativeMomentum: string;
};
confidenceLevel: 'HIGH' | 'MEDIUM' | 'LOW';
dataFreshness: string;
sourceAttribution: string;
methodology: string;
dataQualityWarnings: string[];
forecastBasis: ForecastBasis;
lifecycleCorpus: {
corpusSize: number;
totalObservations: number;
computationTimeMs: number;
};
}
/**
* Default recency window (years) applied to the ACTIVE status filter when no
* explicit dateFrom is provided. Procedures whose best available date predates
* this window cannot be confirmed as currently active and are excluded.
*/
const ACTIVE_RECENCY_YEARS = 10;
/**
* Minimum sample size for a `(type, stage)` cell to be treated as a reliable
* forecasting baseline. Below this the cell is ignored and we fall back to
* the heuristic forecast.
*/
const MIN_SAMPLE_SIZE_FOR_FORECAST = 3;
/**
* Calculate whole-day delta between two date strings (or between a date and
* "now" when `endStr` is omitted). Uses UTC day boundaries with `Math.floor`
* to match {@link import('../utils/lifecycleStatistics.js').daysBetween}, so
* the current-dwell value compared against `p95DwellDays` / `medianDwellDays`
* is computed on the same scale as the corpus statistics.
*/
function daysBetween(dateStr: string, endStr?: string): number {
const start = new Date(dateStr);
const end = endStr !== undefined && endStr !== '' ? new Date(endStr) : new Date();
if (isNaN(start.getTime())) return 0;
Iif (isNaN(end.getTime())) return 0;
const msPerDay = 1000 * 60 * 60 * 24;
const startUtcDays = Math.floor(start.getTime() / msPerDay);
const endUtcDays = Math.floor(end.getTime() / msPerDay);
return Math.max(0, endUtcDays - startUtcDays);
}
/** Classify complexity from days in stage */
function classifyComplexity(days: number): string {
if (days > 60) return 'HIGH';
Iif (days > 30) return 'MEDIUM';
return 'LOW';
}
/**
* Classify bottleneck risk by comparing the current dwell to the historical
* distribution for `(procedureType, stage)`. Falls back to the legacy
* heuristic when no statistics are available so the tool degrades gracefully.
*/
function classifyBottleneckRisk(
days: number,
stats: StageDwellStatistics | undefined,
isStalledHeuristic: boolean
): string {
if (stats !== undefined && stats.sampleSize > 0) {
Eif (days >= stats.p95DwellDays) return 'HIGH';
if (days >= stats.medianDwellDays) return 'MEDIUM';
return 'LOW';
}
if (isStalledHeuristic) return 'HIGH';
if (days > 45) return 'MEDIUM';
return 'LOW';
}
/** Classify bottleneck severity */
function classifyBottleneckSeverity(count: number): string {
Iif (count > 3) return 'CRITICAL';
Iif (count > 1) return 'HIGH';
return 'MODERATE';
}
/** Classify legislative momentum */
function classifyMomentum(healthScore: number): string {
if (healthScore > 80) return 'STRONG';
if (healthScore > 60) return 'MODERATE';
Iif (healthScore > 40) return 'SLOW';
return 'STALLED';
}
/**
* Check if a procedure status indicates completion.
*/
function isStatusCompleted(status: string): boolean {
const lower = status.toLowerCase();
return lower.includes('adopted') || lower.includes('completed') ||
lower.includes('rejected') || lower.includes('withdrawn') || lower.includes('closed');
}
/**
* Compute the dwell time at the current stage from the latest event date.
* Falls back to the procedure's `dateLastActivity` (then `dateInitiated`)
* when no events are available so behavior remains defined when the
* `/events` endpoint is unreachable for a given procedure.
*/
function computeDwellFromEvents(
proc: Procedure,
sortedEvents: readonly EPEvent[]
): { daysInStage: number; latestEventDate: string; latestStage: string } {
if (sortedEvents.length > 0) {
const latest = sortedEvents[sortedEvents.length - 1];
Eif (latest !== undefined) {
return {
daysInStage: daysBetween(latest.date),
latestEventDate: latest.date,
latestStage: normalizeStageKey(latest.type),
};
}
}
const fallback = proc.dateLastActivity !== '' ? proc.dateLastActivity : proc.dateInitiated;
return {
daysInStage: daysBetween(fallback),
latestEventDate: fallback,
latestStage: '',
};
}
/**
* Forecast the remaining days until completion, preferring the historical
* median for the matching `(procedureType, stage)` cell.
*/
function forecastRemainingDays(
isCompleted: boolean,
daysInStage: number,
stats: StageDwellStatistics | undefined
): { estimatedDays: number; basis: ForecastBasis } {
if (isCompleted) return { estimatedDays: 0, basis: 'NOT_APPLICABLE' };
if (stats !== undefined && stats.sampleSize >= MIN_SAMPLE_SIZE_FOR_FORECAST) {
const remaining = Math.max(0, stats.medianRemainingDays - daysInStage);
return { estimatedDays: remaining, basis: 'HISTORICAL_MEDIAN' };
}
return { estimatedDays: Math.max(30, daysInStage * 2), basis: 'INSUFFICIENT_DATA' };
}
/**
* Compute progress metrics for a procedure using the real event timeline.
*/
function computePipelineMetrics(
proc: Procedure,
sortedEvents: readonly EPEvent[],
stats: StageDwellStatistics | undefined
): {
daysInStage: number; isCompleted: boolean; isStalled: boolean;
totalDays: number; progressEstimate: number; velocityScore: number;
estimatedDays: number; basis: ForecastBasis; latestStage: string; latestEventDate: string;
} {
const dwell = computeDwellFromEvents(proc, sortedEvents);
const isCompleted = isStatusCompleted(proc.status);
const isStalledByStats = stats !== undefined && stats.sampleSize > 0
? dwell.daysInStage >= stats.p95DwellDays
: false;
const isStalledByHeuristic = !isCompleted && dwell.daysInStage > 60;
const isStalled = !isCompleted && (isStalledByStats || isStalledByHeuristic);
const initiated = proc.dateInitiated !== '' ? proc.dateInitiated : '';
const lastAct = proc.dateLastActivity !== '' ? proc.dateLastActivity : undefined;
const totalDays = daysBetween(initiated, lastAct);
const progressEstimate = isCompleted ? 100 : Math.min(90, Math.max(5, Math.round(totalDays / 10)));
const velocityScore = isStalled ? 20 : Math.min(100, 100 - Math.min(80, dwell.daysInStage));
const { estimatedDays, basis } = forecastRemainingDays(isCompleted, dwell.daysInStage, stats);
return {
daysInStage: dwell.daysInStage,
isCompleted,
isStalled,
totalDays,
progressEstimate,
velocityScore,
estimatedDays,
basis,
latestStage: dwell.latestStage,
latestEventDate: dwell.latestEventDate,
};
}
/**
* Build the lifecycleEvents array echoed back to callers for traceability.
* The events are returned in chronological order, normalised but preserving
* the original raw type for downstream consumers.
*/
function toLifecycleEvents(sortedEvents: readonly EPEvent[]): PipelineLifecycleEvent[] {
return sortedEvents.map((e) => ({
date: e.date,
stage: normalizeStageKey(e.type),
rawType: e.type,
title: e.title,
}));
}
/**
* Resolve the procedure's display stage, preferring the latest event stage,
* then the procedure's metadata stage/status fields. Falls back to `'Unknown'`
* when no stage signal is available.
*/
function resolveCurrentStage(latestStage: string, proc: Procedure): string {
if (latestStage !== '') return latestStage;
if (proc.stage !== '') return proc.stage;
Iif (proc.status !== '') return proc.status;
return 'Unknown';
}
/**
* Transform a real EP API Procedure into a PipelineItem, using the procedure's
* lifecycle events when available.
*/
function procedureToPipelineItem(
proc: Procedure,
sortedEvents: readonly EPEvent[],
model: LifecycleStatisticsModel
): { item: PipelineItem; basis: ForecastBasis } {
const procedureType = proc.type !== '' ? proc.type : 'UNKNOWN';
const stageFromEvents = sortedEvents.length > 0
? normalizeStageKey(sortedEvents[sortedEvents.length - 1]?.type ?? '')
: '';
const stats = lookupStageStatistics(model, procedureType, stageFromEvents);
const m = computePipelineMetrics(proc, sortedEvents, stats);
const currentStage = resolveCurrentStage(m.latestStage, proc);
const committee = proc.responsibleCommittee !== '' ? proc.responsibleCommittee : 'Unknown';
const stageLabel = currentStage !== 'Unknown' ? currentStage : 'processing';
const nextAction = m.isCompleted ? 'COMPLETED' : `Continue ${stageLabel}`;
return {
item: {
procedureId: proc.id,
title: proc.title,
type: proc.type,
currentStage,
committee,
daysInCurrentStage: m.daysInStage,
isStalled: m.isStalled,
nextExpectedAction: nextAction,
computedAttributes: {
progressPercentage: m.progressEstimate,
velocityScore: m.velocityScore,
complexityIndicator: classifyComplexity(m.daysInStage),
estimatedCompletionDays: m.estimatedDays,
bottleneckRisk: classifyBottleneckRisk(m.daysInStage, stats, m.isStalled),
},
lifecycleEvents: toLifecycleEvents(sortedEvents),
},
basis: m.basis,
};
}
/**
* Check if a procedure's dates pass the recency cut-off for the ACTIVE filter.
* Returns false if the procedure has no temporal data or is older than the cut-off.
*/
function isWithinRecencyCutoff(
lastActivity: string,
initiated: string | undefined,
cutoffDate: string
): boolean {
if (lastActivity === '' && initiated === undefined) return false;
const referenceDate = lastActivity !== '' ? lastActivity : initiated;
return referenceDate !== undefined && referenceDate >= cutoffDate;
}
/** Check if a procedure matches an explicit date range (dateFrom / dateTo) */
function matchesDateRange(
lastActivity: string,
initiated: string | undefined,
dateFrom: string | undefined,
dateTo: string | undefined
): boolean {
if (dateFrom !== undefined && lastActivity !== '' && lastActivity < dateFrom) return false;
Iif (dateTo !== undefined && initiated !== undefined && initiated > dateTo) return false;
return true;
}
/** Check if item matches status filter */
function matchesStatusFilter(item: PipelineItem, status: string): boolean {
if (status === 'ALL') return true;
if (status === 'ACTIVE') {
return !item.isStalled
&& item.computedAttributes.progressPercentage < 100
&& item.currentStage !== 'Unknown';
}
Eif (status === 'STALLED') return item.isStalled;
if (status === 'COMPLETED') return item.computedAttributes.progressPercentage >= 100;
return true;
}
/** Check if item matches committee filter */
function matchesCommitteeFilter(item: PipelineItem, committee: string | undefined): boolean {
Eif (committee === undefined) return true;
return item.committee === committee;
}
/**
* Compare a stage string for sort stability (named to avoid nested ternaries).
*/
function compareStage(a: string, b: string): number {
Iif (a < b) return -1;
Eif (a > b) return 1;
return 0;
}
/**
* Decide whether a pipeline item should be counted as a bottleneck.
*
* Returns `true` when the procedure's dwell at its current stage is at or
* above the historical 95th percentile, or — when no statistics exist for the
* `(type, stage)` cell — when the legacy stalled heuristic applies.
*/
function isBottleneckItem(
item: PipelineItem,
stats: StageDwellStatistics | undefined,
): boolean {
if (stats !== undefined && stats.sampleSize > 0) {
return item.daysInCurrentStage >= stats.p95DwellDays;
}
return stats === undefined && item.isStalled;
}
interface BottleneckBucket {
count: number;
totalDays: number;
thresholdDays: number;
}
/**
* Detect bottlenecks from pipeline items whose dwell at the current stage is
* at or above the historical 95th percentile. Procedures with no matching
* statistics fall back to the legacy `isStalled` heuristic so the tool keeps
* producing a useful list even when the corpus is empty.
*
* **Cross-type aggregation:** bottlenecks are keyed by **stage label only**
* (e.g. `"REFERRAL"`), not `(type, stage)`. This is intentional — callers
* looking at "which stage is stuck right now" want a single view across
* procedure types. The per-procedure `bottleneckRisk` field (and the
* underlying p95 cell) remains type-aware, so the precise threshold for an
* individual procedure is unaffected by cross-type aggregation here.
* `thresholdDays` reports the **maximum** p95 seen across the contributing
* `(type, stage)` cells — interpret it as a conservative high-water mark
* for the worst type in that stage bucket.
*/
function detectBottlenecks(pipeline: PipelineItem[], model: LifecycleStatisticsModel): BottleneckInfo[] {
const stageBuckets: Record<string, BottleneckBucket> = {};
for (const item of pipeline) {
const stats = lookupStageStatistics(model, item.type !== '' ? item.type : 'UNKNOWN', item.currentStage);
if (!isBottleneckItem(item, stats)) continue;
const entry = stageBuckets[item.currentStage] ?? {
count: 0,
totalDays: 0,
thresholdDays: stats?.p95DwellDays ?? 0,
};
entry.count++;
entry.totalDays += item.daysInCurrentStage;
Iif (stats !== undefined && stats.p95DwellDays > entry.thresholdDays) {
entry.thresholdDays = stats.p95DwellDays;
}
stageBuckets[item.currentStage] = entry;
}
return Object.entries(stageBuckets)
.map(([stage, data]) => ({
stage,
procedureCount: data.count,
avgDaysStuck: Math.round(data.totalDays / data.count),
severity: classifyBottleneckSeverity(data.count),
thresholdDays: data.thresholdDays,
}))
.sort((a, b) => {
Iif (b.procedureCount !== a.procedureCount) return b.procedureCount - a.procedureCount;
return compareStage(a.stage, b.stage);
});
}
/** Compute pipeline summary statistics */
function computePipelineSummary(pipeline: PipelineItem[]): {
activeCount: number; stalledCount: number; completedCount: number; avgDays: number;
} {
const activeCount = pipeline.filter(p => !p.isStalled && p.computedAttributes.progressPercentage < 100).length;
const stalledCount = pipeline.filter(p => p.isStalled).length;
const completedCount = pipeline.filter(p => p.computedAttributes.progressPercentage >= 100).length;
const totalDays = pipeline.reduce((sum, p) => sum + p.daysInCurrentStage, 0);
const avgDays = pipeline.length > 0 ? Math.round(totalDays / pipeline.length) : 0;
return { activeCount, stalledCount, completedCount, avgDays };
}
/** Compute pipeline health metrics */
function computeHealthMetrics(pipeline: PipelineItem[], summary: ReturnType<typeof computePipelineSummary>): {
healthScore: number; throughputRate: number; stalledRate: number;
} {
const stalledRate = pipeline.length > 0 ? summary.stalledCount / pipeline.length : 0;
const healthScore = Math.round((1 - stalledRate) * 100 * 100) / 100;
const throughputRate = pipeline.length > 0
? Math.round((summary.completedCount / pipeline.length) * 100 * 100) / 100
: 0;
return { healthScore, throughputRate, stalledRate };
}
/**
* Aggregate the per-procedure forecast bases into a single envelope-level
* basis. If any procedure used a historical median forecast we report
* `HISTORICAL_MEDIAN`; otherwise we report `INSUFFICIENT_DATA`.
*/
function aggregateForecastBasis(perItemBases: readonly ForecastBasis[]): ForecastBasis {
// Exclude NOT_APPLICABLE (completed procedures) from the envelope-level determination
const actionable = perItemBases.filter((b) => b !== 'NOT_APPLICABLE');
if (actionable.length === 0) return 'NOT_APPLICABLE';
return actionable.some((b) => b === 'HISTORICAL_MEDIAN')
? 'HISTORICAL_MEDIAN'
: 'INSUFFICIENT_DATA';
}
/**
* Build the human-readable list of data-quality warnings appended to the
* response envelope.
*/
function buildWarnings(
pipelineSize: number,
envelopeBasis: ForecastBasis,
unknownEnrichmentCount: number,
eventFetchFailures: number,
): string[] {
const warnings: string[] = [];
Eif (pipelineSize < 10) {
warnings.push('Small procedure sample (< 10) — pipeline health metrics may not be statistically representative');
}
if (unknownEnrichmentCount > 0) {
warnings.push(`${String(unknownEnrichmentCount)} procedure(s) excluded from ACTIVE filter due to missing enrichment data (stage/committee unknown) — these may be historical or incomplete records`);
}
if (eventFetchFailures > 0) {
warnings.push(`${String(eventFetchFailures)} procedure(s) had no lifecycle events available — fell back to procedure-metadata dates for dwell computation`);
}
if (envelopeBasis === 'INSUFFICIENT_DATA') {
warnings.push('Forecast basis: insufficient historical lifecycle data — `estimatedCompletionDays` uses a heuristic fallback');
}
return warnings;
}
interface BuildAnalysisInput {
reportFrom: string;
reportTo: string;
params: ReturnType<typeof MonitorLegislativePipelineSchema.parse>;
pipeline: PipelineItem[];
model: LifecycleStatisticsModel;
envelopeBasis: ForecastBasis;
unknownEnrichmentCount: number;
eventFetchFailures: number;
}
/**
* Build the analysis envelope.
*/
function buildAnalysis(input: BuildAnalysisInput): LegislativePipelineAnalysis {
const { reportFrom, reportTo, params, pipeline, model, envelopeBasis,
unknownEnrichmentCount, eventFetchFailures } = input;
const summary = computePipelineSummary(pipeline);
const bottlenecks = detectBottlenecks(pipeline, model);
const health = computeHealthMetrics(pipeline, summary);
const warnings = buildWarnings(pipeline.length, envelopeBasis, unknownEnrichmentCount, eventFetchFailures);
return {
period: { from: reportFrom, to: reportTo },
filter: { ...(params.committee !== undefined ? { committee: params.committee } : {}), status: params.status },
pipeline,
summary: {
totalProcedures: pipeline.length,
activeCount: summary.activeCount,
stalledCount: summary.stalledCount,
completedCount: summary.completedCount,
avgDaysInPipeline: summary.avgDays,
},
bottlenecks,
computedAttributes: {
pipelineHealthScore: health.healthScore,
throughputRate: health.throughputRate,
bottleneckIndex: Math.round(health.stalledRate * summary.avgDays * 100) / 100,
stalledProcedureRate: Math.round(health.stalledRate * 100 * 100) / 100,
estimatedClearanceTime: summary.avgDays * Math.max(1, summary.activeCount),
legislativeMomentum: classifyMomentum(health.healthScore),
},
confidenceLevel: pipeline.length >= 10 && envelopeBasis === 'HISTORICAL_MEDIAN'
? 'MEDIUM'
: 'LOW',
dataFreshness: 'Real-time EP API data — procedures from /procedures and lifecycle from /procedures/{id}/events',
sourceAttribution: 'European Parliament Open Data Portal - data.europarl.europa.eu',
methodology: 'Lifecycle-driven pipeline analysis using EP API /procedures and /procedures/{id}/events. '
+ 'For each procedure in scope the authoritative event sequence is fetched in parallel (bounded ≤8). '
+ '`daysInCurrentStage` is the delta between the most recent event and now. '
+ '`bottleneckRisk` is a percentile bucket of the current dwell against the historical distribution '
+ 'for the same (procedureType, stage) observed across the latest 500 procedures (30 min cache). '
+ '`estimatedCompletionDays` uses the historical median remaining-time at the current stage; '
+ 'when no comparable historical sample exists the response is flagged `INSUFFICIENT_DATA`. '
+ 'Bottlenecks are aggregated from procedures whose dwell is ≥ the 95th percentile.',
dataQualityWarnings: warnings,
forecastBasis: envelopeBasis,
lifecycleCorpus: {
corpusSize: model.corpusSize,
totalObservations: model.totalObservations,
computationTimeMs: model.computationTimeMs,
},
};
}
interface OperationContext {
params: ReturnType<typeof MonitorLegislativePipelineSchema.parse>;
reportFrom: string;
reportTo: string;
activeCutoffDate: string | undefined;
}
/**
* Resolve the reporting period and ACTIVE recency cut-off from input params.
*/
function resolveOperationContext(
params: ReturnType<typeof MonitorLegislativePipelineSchema.parse>
): OperationContext {
const toIsoDate = (d: Date): string => d.toISOString().slice(0, 10);
const todayIso = toIsoDate(new Date());
const defaultFromIso = ((): string => {
const d = new Date();
d.setUTCDate(d.getUTCDate() - 30);
return toIsoDate(d);
})();
const reportFrom = params.dateFrom ?? defaultFromIso;
const reportTo = params.dateTo ?? todayIso;
const activeCutoffDate = ((): string | undefined => {
if (params.status !== 'ACTIVE' || params.dateFrom !== undefined) return undefined;
const referenceYear = parseInt(
(params.dateTo ?? todayIso).slice(0, 4),
10
);
return `${String(referenceYear - ACTIVE_RECENCY_YEARS)}-01-01`;
})();
return { params, reportFrom, reportTo, activeCutoffDate };
}
/**
* Apply the recency + explicit date-range filters to the raw procedure list.
*/
function applyDateFilters(
procedures: readonly Procedure[],
ctx: OperationContext,
): Procedure[] {
return procedures.filter(proc => {
const lastActivity = proc.dateLastActivity !== '' ? proc.dateLastActivity : proc.dateInitiated;
const initiated = proc.dateInitiated !== '' ? proc.dateInitiated : undefined;
if (ctx.activeCutoffDate !== undefined && !isWithinRecencyCutoff(lastActivity, initiated, ctx.activeCutoffDate)) {
return false;
}
return matchesDateRange(lastActivity, initiated, ctx.params.dateFrom, ctx.params.dateTo);
});
}
/**
* Build per-procedure pipeline items and count event-fetch failures.
*
* Returns the forecast basis indexed by `procedureId` so downstream
* filtering can correlate items to their basis without relying on
* positional indices (which become misaligned after filtering).
*/
function buildPipelineItems(
filteredProcs: readonly Procedure[],
eventsByProcedureId: ReadonlyMap<string, EPEvent[]>,
model: LifecycleStatisticsModel,
): { items: PipelineItem[]; basisByProcedureId: Map<string, ForecastBasis>; eventFetchFailures: number } {
const items: PipelineItem[] = [];
const basisByProcedureId = new Map<string, ForecastBasis>();
let eventFetchFailures = 0;
for (const proc of filteredProcs) {
const rawEvents = eventsByProcedureId.get(proc.id);
if (rawEvents === undefined || rawEvents.length === 0) {
eventFetchFailures++;
}
const sortedEvents = sortEventsChronologically(rawEvents ?? []);
const { item, basis } = procedureToPipelineItem(proc, sortedEvents, model);
items.push(item);
basisByProcedureId.set(proc.id, basis);
}
return { items, basisByProcedureId, eventFetchFailures };
}
/**
* Load the lifecycle statistics model **from cache only** — never rebuild on
* the request path.
*
* The corpus build (`/procedures` + up to {@link CORPUS_SIZE}
* `/procedures/{id}/events`) cannot reliably finish inside this tool's
* {@link OPERATION_TIMEOUT_MS} envelope on a cold runner: even with the
* cooperative `deadline`, in-flight EP API requests cannot be cancelled and
* the budget's worth of token-bucket spend leaves the request's own
* `/procedures` + `fetchEventsBounded` starved by the 100 req/min rate
* limit. The visible symptom is the outer 30 s `withTimeout` firing and the
* tool returning a {@link buildTimeoutResponse} envelope without
* `pipeline`/`summary`.
*
* On a cache miss we therefore return {@link emptyLifecycleStatisticsModel}
* synchronously and degrade per-procedure forecasts to `INSUFFICIENT_DATA`.
* The corpus model is warmed out-of-band (e.g. by sibling tools or a
* scheduled job) via {@link getLifecycleStatistics}; once cached, subsequent
* requests pick it up immediately. This preserves the response shape on
* cold integration runs while leaving the warm-path forecasts unchanged.
*/
function loadCachedLifecycleModel(): LifecycleStatisticsModel {
const cached = getCachedLifecycleStatistics();
Eif (cached !== null) {
return cached;
}
// Audit-friendly fallback so cold-runner runs are observable without
// emitting a noisy error: per-procedure forecasts degrade to
// `INSUFFICIENT_DATA` and the response envelope is unchanged.
console.error(
'[monitor_legislative_pipeline] Lifecycle cache miss — using empty model;'
+ ' per-procedure forecasts will degrade to INSUFFICIENT_DATA',
);
return emptyLifecycleStatisticsModel();
}
/**
* Run the core pipeline-monitoring operation: load procedures + lifecycle
* statistics, fetch per-procedure events, score each procedure, filter, and
* assemble the response envelope.
*/
async function runPipelineOperation(
params: ReturnType<typeof MonitorLegislativePipelineSchema.parse>
): Promise<LegislativePipelineAnalysis> {
const ctx = resolveOperationContext(params);
// Load the lifecycle model FROM CACHE ONLY. The corpus rebuild is far too
// expensive for the request path (see `loadCachedLifecycleModel` JSDoc);
// cache misses degrade per-procedure forecasts to `INSUFFICIENT_DATA` but
// keep the response envelope intact and free the rate-limit budget for
// the request's own `/procedures` + `fetchEventsBounded` calls.
const model = loadCachedLifecycleModel();
const procedures = await epClient.getProcedures({ limit: params.limit });
const filteredProcs = applyDateFilters(procedures.data, ctx);
const eventsByProcedureId = await fetchEventsBounded(filteredProcs);
const { items: allMappedItems, basisByProcedureId, eventFetchFailures } =
buildPipelineItems(filteredProcs, eventsByProcedureId, model);
const unknownEnrichmentCount = params.status === 'ACTIVE'
? allMappedItems.filter(item => item.currentStage === 'Unknown').length
: 0;
const allItems = allMappedItems
.filter((item) => matchesStatusFilter(item, params.status))
.filter((item) => matchesCommitteeFilter(item, params.committee));
const pipeline = allItems.slice(0, params.limit);
const visibleBases: ForecastBasis[] = pipeline
.map((item) => basisByProcedureId.get(item.procedureId))
.filter((b): b is ForecastBasis => b !== undefined);
const envelopeBasis = aggregateForecastBasis(visibleBases);
return buildAnalysis({
reportFrom: ctx.reportFrom,
reportTo: ctx.reportTo,
params,
pipeline,
model,
envelopeBasis,
unknownEnrichmentCount,
eventFetchFailures,
});
}
/**
* Handles the monitor_legislative_pipeline MCP tool request.
*
* Monitors the European Parliament's active legislative pipeline by fetching
* real procedures and their authoritative event timelines from the EP API
* (`/procedures` + `/procedures/{id}/events`) and computing lifecycle-driven
* health metrics: percentile-based bottleneck detection, historical-median
* completion forecasts, and stage-aware dwell statistics.
*
* @param args - Raw tool arguments, validated against {@link MonitorLegislativePipelineSchema}
* @returns MCP tool result containing pipeline items with stage, lifecycle events,
* forecast basis, summary counts, detected bottlenecks (≥95th percentile dwell),
* pipeline health score, throughput rate, bottleneck index, and legislative momentum.
* @throws - If `args` fails schema validation (e.g., missing required fields or invalid format)
* - If the European Parliament API is unreachable for the primary procedure list
*
* @example
* ```typescript
* const result = await handleMonitorLegislativePipeline({
* status: 'ACTIVE',
* committee: 'ENVI',
* dateFrom: '2024-01-01',
* dateTo: '2024-12-31',
* limit: 20
* });
* // Returns pipeline health score, stalled/active/completed counts,
* // bottleneck list, lifecycleEvents per procedure, and forecastBasis.
* ```
*
* @security - Input is validated with Zod before any API call.
* - Personal data in responses is minimised per GDPR Article 5(1)(c).
* - Bounded concurrency (≤8 parallel) on the event fan-out limits API load.
* - Lifecycle corpus is cached 30 min; only event types/dates are retained.
* - All requests are rate-limited and audit-logged per ISMS Policy AU-002.
* @since 0.8.0
* @see {@link monitorLegislativePipelineToolMetadata} for MCP schema registration
* @see {@link handleTrackLegislation} for individual procedure stage and timeline tracking
*/
export async function handleMonitorLegislativePipeline(
args: unknown
): Promise<ToolResult> {
const params = MonitorLegislativePipelineSchema.parse(args);
try {
const analysis = await withTimeout(
runPipelineOperation(params),
OPERATION_TIMEOUT_MS,
'monitor_legislative_pipeline operation timed out'
);
return { content: [{ type: 'text', text: JSON.stringify(analysis, null, 2) }] };
} catch (error: unknown) {
Iif (error instanceof TimeoutError) {
// Return a *shape-stable* timeout envelope so downstream consumers can
// still introspect `pipeline` / `summary` / `methodology` (matching the
// happy-path schema) and surface the timeout via `dataQualityWarnings`
// + LOW confidence rather than the generic `buildTimeoutResponse` which
// drops the response shape entirely.
return buildPipelineTimeoutResponse(params);
}
// Preserve the original error as `cause` for server-side diagnostics
// (logs/audit) while keeping the client-visible message generic so we
// don't leak upstream URLs, status text, or other internal details.
throw new ToolError({
toolName: 'monitor_legislative_pipeline',
operation: 'runPipelineOperation',
message: 'Failed to monitor legislative pipeline',
cause: error,
isRetryable: true,
});
}
}
/**
* Build a {@link LegislativePipelineAnalysis}-shaped envelope when the outer
* operation exceeded {@link OPERATION_TIMEOUT_MS}. Preserves the response
* contract — `pipeline`, `summary`, `methodology` (with the `EP API` token
* sibling tools/tests assert on) — while clearly marking the call as a
* partial / unavailable result via {@link LegislativePipelineAnalysis.confidenceLevel}
* `'LOW'`, an explicit `dataQualityWarnings` entry, and `forecastBasis:
* 'INSUFFICIENT_DATA'`. Used to keep integration tests deterministic when
* the EP API is slow/rate-limited and the cooperative deadline fires.
*
* @internal
*/
function buildPipelineTimeoutResponse(
params: ReturnType<typeof MonitorLegislativePipelineSchema.parse>
): ToolResult {
const ctx = resolveOperationContext(params);
const timeoutWarning =
`monitor_legislative_pipeline timed out after ${String(OPERATION_TIMEOUT_MS)}ms `
+ '— EP API was slow or rate-limited. Retry with a narrower committee/date filter '
+ 'or smaller `limit`; the lifecycle corpus may also still be warming.';
const analysis: LegislativePipelineAnalysis = {
period: { from: ctx.reportFrom, to: ctx.reportTo },
filter: {
...(params.committee !== undefined ? { committee: params.committee } : {}),
status: params.status,
},
pipeline: [],
summary: {
totalProcedures: 0,
activeCount: 0,
stalledCount: 0,
completedCount: 0,
avgDaysInPipeline: 0,
},
bottlenecks: [],
computedAttributes: {
pipelineHealthScore: 0,
throughputRate: 0,
bottleneckIndex: 0,
stalledProcedureRate: 0,
estimatedClearanceTime: 0,
legislativeMomentum: 'UNKNOWN',
},
confidenceLevel: 'LOW',
dataFreshness: 'Timeout — EP API was unreachable within the configured window',
sourceAttribution: 'European Parliament Open Data Portal - data.europarl.europa.eu',
methodology: 'Lifecycle-driven pipeline analysis using EP API /procedures and '
+ '/procedures/{id}/events. Operation exceeded the configured timeout before any '
+ 'procedures could be scored — no analytical conclusions were produced. '
+ 'Retry with a narrower filter or smaller `limit`.',
dataQualityWarnings: [timeoutWarning],
forecastBasis: 'INSUFFICIENT_DATA',
lifecycleCorpus: { corpusSize: 0, totalObservations: 0, computationTimeMs: 0 },
};
return { content: [{ type: 'text', text: JSON.stringify(analysis, null, 2) }] };
}
/**
* Tool metadata for MCP registration
*/
export const monitorLegislativePipelineToolMetadata = {
name: 'monitor_legislative_pipeline',
description: 'Monitor legislative pipeline status with lifecycle-driven bottleneck detection and timeline forecasting. Tracks procedures through their authoritative event sequence (REFERRAL → COM_VOTE → EP_ADOPTION → SIGNATURE / REJECTION). Returns pipeline health score, throughput rate, bottleneck index (procedures with dwell ≥ 95th percentile of historical distribution), stalled procedure rate, legislative momentum, per-procedure lifecycleEvents, and a forecastBasis discriminator (HISTORICAL_MEDIAN | INSUFFICIENT_DATA | NOT_APPLICABLE — the last when every visible procedure is already completed).',
inputSchema: {
type: 'object' as const,
properties: {
committee: {
type: 'string',
description: 'Filter by committee',
minLength: 1,
maxLength: 100
},
status: {
type: 'string',
enum: ['ALL', 'ACTIVE', 'STALLED', 'COMPLETED'],
description: 'Pipeline status filter',
default: 'ACTIVE'
},
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}$'
},
limit: {
type: 'number',
description: 'Maximum results to return (1-100)',
minimum: 1,
maximum: 100,
default: 20
}
}
}
};
|