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 | 72x 72x 72x 72x 72x 72x 3x 3x 72x 2x 2x 2x 72x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 5x 2x 2x 2x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 54x 97x 97x 13x 84x 10x 74x 10x 64x 33x 33x 13x 10x 10x 72x 72x 72x 873x 873x 22x 7x 7x 7x 7x 7x 1x 1x 1x 6x 1x 1x 1x 5x 1x 1x 1x 19x 4x 4x 4x 4x 4x 4x 8x 4x 15x 50x 72x 72x 72x 286x 26x 46x 72x 72x 72x 72x 12x 72x 72x 1x 72x 72x 1x 72x 97x 97x 97x 97x 97x 8x 4x 3x 3x 1x 1x 93x 93x 129x 89x 93x 89x 97x 25x 25x 25x 72x 72x 72x 72x 26x 26x 72x 72x 72x 278x 278x 164x 278x 680x 24x 24x 24x 24x 72x 27x 6x 6x 6x 72x 42x 42x 42x 72x 72x 72x 72x 14x 14x 10x 10x 10x 72x 92x 258x 258x 196x 258x 700x 6x 3x 3x 3x 3x 72x 13x 72x 72x 64x 22x 21x 21x 21x 21x 21x 21x 1x 72x 72x 12x 12x 60x | /**
* AI Query Suggestion Engine
* Parse natural language intent and generate query structure
*/
import type { CubeMetadata } from '../types/metadata'
import type { SemanticQuery, TimeDimension } from '../types/query'
import { discoverCubes, findBestFieldMatch } from './discovery'
/**
* Suggested query result
*/
export interface QuerySuggestion {
query: Partial<SemanticQuery>
confidence: number
reasoning: string[]
warnings?: string[]
/** Detected analysis mode */
analysisMode: 'query' | 'funnel' | 'flow' | 'retention'
/** Next steps when mode != 'query' */
nextSteps?: string[]
}
/**
* Time expression patterns
*/
interface TimeExpression {
pattern: RegExp
getDateRange: () => [string, string]
granularity?: string
}
/**
* Get time expressions with current date context
*/
function getTimeExpressions(): TimeExpression[] {
const now = new Date()
const today = now.toISOString().split('T')[0]
// Helper to format date
const formatDate = (d: Date): string => d.toISOString().split('T')[0]
// Helper to get start of period
const startOfMonth = (d: Date): Date => new Date(d.getFullYear(), d.getMonth(), 1)
const startOfYear = (d: Date): Date => new Date(d.getFullYear(), 0, 1)
const startOfQuarter = (d: Date): Date => {
const quarter = Math.floor(d.getMonth() / 3)
return new Date(d.getFullYear(), quarter * 3, 1)
}
const startOfWeek = (d: Date): Date => {
const day = d.getDay()
const diff = d.getDate() - day + (day === 0 ? -6 : 1)
return new Date(d.getFullYear(), d.getMonth(), diff)
}
return [
// Today
{
pattern: /\btoday\b/i,
getDateRange: () => [today, today],
granularity: 'day'
},
// Yesterday
{
pattern: /\byesterday\b/i,
getDateRange: () => {
const yesterday = new Date(now)
yesterday.setDate(yesterday.getDate() - 1)
const d = formatDate(yesterday)
return [d, d]
},
granularity: 'day'
},
// This week
{
pattern: /\bthis week\b/i,
getDateRange: () => [formatDate(startOfWeek(now)), today],
granularity: 'day'
},
// Last week
{
pattern: /\blast week\b/i,
getDateRange: () => {
const lastWeekStart = new Date(startOfWeek(now))
lastWeekStart.setDate(lastWeekStart.getDate() - 7)
const lastWeekEnd = new Date(lastWeekStart)
lastWeekEnd.setDate(lastWeekEnd.getDate() + 6)
return [formatDate(lastWeekStart), formatDate(lastWeekEnd)]
},
granularity: 'day'
},
// This month
{
pattern: /\bthis month\b/i,
getDateRange: () => [formatDate(startOfMonth(now)), today],
granularity: 'day'
},
// Last month
{
pattern: /\blast month\b/i,
getDateRange: () => {
const lastMonth = new Date(now.getFullYear(), now.getMonth() - 1, 1)
const lastMonthEnd = new Date(now.getFullYear(), now.getMonth(), 0)
return [formatDate(lastMonth), formatDate(lastMonthEnd)]
},
granularity: 'day'
},
// This quarter
{
pattern: /\bthis quarter\b/i,
getDateRange: () => [formatDate(startOfQuarter(now)), today],
granularity: 'month'
},
// Last quarter
{
pattern: /\blast quarter\b/i,
getDateRange: () => {
const lastQuarterStart = new Date(startOfQuarter(now))
lastQuarterStart.setMonth(lastQuarterStart.getMonth() - 3)
const lastQuarterEnd = new Date(startOfQuarter(now))
lastQuarterEnd.setDate(lastQuarterEnd.getDate() - 1)
return [formatDate(lastQuarterStart), formatDate(lastQuarterEnd)]
},
granularity: 'month'
},
// This year
{
pattern: /\bthis year\b/i,
getDateRange: () => [formatDate(startOfYear(now)), today],
granularity: 'month'
},
// Last year
{
pattern: /\blast year\b/i,
getDateRange: () => {
const lastYear = new Date(now.getFullYear() - 1, 0, 1)
const lastYearEnd = new Date(now.getFullYear() - 1, 11, 31)
return [formatDate(lastYear), formatDate(lastYearEnd)]
},
granularity: 'month'
},
// Last N days
{
pattern: /\blast (\d+) days?\b/i,
getDateRange: () => {
// Default to 7 days, actual value extracted separately
const start = new Date(now)
start.setDate(start.getDate() - 7)
return [formatDate(start), today]
},
granularity: 'day'
},
// Last N weeks
{
pattern: /\blast (\d+) weeks?\b/i,
getDateRange: () => {
const start = new Date(now)
start.setDate(start.getDate() - 7 * 4)
return [formatDate(start), today]
},
granularity: 'week'
},
// Last N months
{
pattern: /\blast (\d+) months?\b/i,
getDateRange: () => {
const start = new Date(now)
start.setMonth(start.getMonth() - 3)
return [formatDate(start), today]
},
granularity: 'month'
},
// Q1, Q2, Q3, Q4 patterns
{
pattern: /\bq([1-4])\b/i,
getDateRange: () => {
// Default Q1 of current year
return [formatDate(new Date(now.getFullYear(), 0, 1)), formatDate(new Date(now.getFullYear(), 2, 31))]
},
granularity: 'month'
}
]
}
/**
* Analysis mode detection patterns
*/
const ANALYSIS_MODE_PATTERNS = {
funnel: /\b(funnel|conversion|drop.?off|steps?|journey|pipeline|stages?)\b/i,
flow: /\b(flows?|paths?|sequence|before|after|next|previous|user.?journey)\b/i,
retention: /\b(retention|cohort|return|churn|comeback|retained|day.?\d+)\b/i
}
/**
* Detect analysis mode from natural language
*/
function detectAnalysisMode(text: string): 'query' | 'funnel' | 'flow' | 'retention' {
const lowerText = text.toLowerCase()
if (ANALYSIS_MODE_PATTERNS.funnel.test(lowerText)) {
return 'funnel'
}
if (ANALYSIS_MODE_PATTERNS.flow.test(lowerText)) {
return 'flow'
}
if (ANALYSIS_MODE_PATTERNS.retention.test(lowerText)) {
return 'retention'
}
return 'query'
}
/**
* Generate next steps for analysis modes
*/
function generateNextSteps(mode: 'funnel' | 'flow' | 'retention', cubeName?: string): string[] {
const cubeRef = cubeName ? cubeName : 'the relevant cube'
switch (mode) {
case 'funnel':
return [
`Use /mcp/discover to get ${cubeRef} funnel configuration and schema`,
`Query the event dimension to discover available event types for funnel steps`,
'Build funnel query with discovered values using the schema from discover'
]
case 'flow':
return [
`Use /mcp/discover to get ${cubeRef} flow configuration and schema`,
`Query the event dimension to discover available event types`,
'Build flow query specifying the starting event and steps before/after'
]
case 'retention':
return [
`Use /mcp/discover to get ${cubeRef} retention configuration and schema`,
'Build retention query specifying granularity (day/week/month) and number of periods'
]
}
}
/**
* Parse time expression from natural language
*/
function parseTimeExpression(text: string): { dateRange: [string, string]; granularity?: string } | null {
const expressions = getTimeExpressions()
const lowerText = text.toLowerCase()
for (const expr of expressions) {
const match = lowerText.match(expr.pattern)
if (match) {
// Handle patterns with numeric capture groups
if (match[1] && /^\d+$/.test(match[1])) {
const n = parseInt(match[1], 10)
const now = new Date()
const today = now.toISOString().split('T')[0]
const formatDate = (d: Date): string => d.toISOString().split('T')[0]
if (/days?/.test(lowerText)) {
const start = new Date(now)
start.setDate(start.getDate() - n)
return { dateRange: [formatDate(start), today], granularity: 'day' }
}
if (/weeks?/.test(lowerText)) {
const start = new Date(now)
start.setDate(start.getDate() - n * 7)
return { dateRange: [formatDate(start), today], granularity: n <= 4 ? 'day' : 'week' }
}
if (/months?/.test(lowerText)) {
const start = new Date(now)
start.setMonth(start.getMonth() - n)
return { dateRange: [formatDate(start), today], granularity: n <= 3 ? 'day' : 'month' }
}
}
// Handle Q1-Q4
if (/^q[1-4]$/i.test(match[0])) {
const quarter = parseInt(match[1], 10)
const now = new Date()
const year = now.getFullYear()
const startMonth = (quarter - 1) * 3
const start = new Date(year, startMonth, 1)
const end = new Date(year, startMonth + 3, 0)
const formatDate = (d: Date): string => d.toISOString().split('T')[0]
return { dateRange: [formatDate(start), formatDate(end)], granularity: 'month' }
}
return { dateRange: expr.getDateRange(), granularity: expr.granularity }
}
}
return null
}
/**
* Detect aggregation intent from natural language
*/
function detectAggregationIntent(text: string): { type: 'sum' | 'count' | 'avg' | 'max' | 'min'; confidence: number } | null {
const lowerText = text.toLowerCase()
const patterns: Array<{ pattern: RegExp; type: 'sum' | 'count' | 'avg' | 'max' | 'min' }> = [
{ pattern: /\b(total|sum|combined)\b/i, type: 'sum' },
{ pattern: /\b(count|number of|how many)\b/i, type: 'count' },
{ pattern: /\b(average|avg|mean)\b/i, type: 'avg' },
{ pattern: /\b(maximum|max|highest|top)\b/i, type: 'max' },
{ pattern: /\b(minimum|min|lowest|bottom)\b/i, type: 'min' }
]
for (const { pattern, type } of patterns) {
if (pattern.test(lowerText)) {
return { type, confidence: 0.8 }
}
}
return null
}
/**
* Detect grouping/breakdown intent
*/
function detectGroupingIntent(text: string): string[] {
const lowerText = text.toLowerCase()
const groupingKeywords: string[] = []
// Look for "by X" patterns
const byPattern = /\bby\s+(\w+(?:\s+\w+)?)/gi
let match
while ((match = byPattern.exec(lowerText)) !== null) {
groupingKeywords.push(match[1].trim())
}
// Look for "per X" patterns
const perPattern = /\bper\s+(\w+)/gi
while ((match = perPattern.exec(lowerText)) !== null) {
groupingKeywords.push(match[1].trim())
}
// Look for "for each X" patterns
const forEachPattern = /\bfor each\s+(\w+)/gi
while ((match = forEachPattern.exec(lowerText)) !== null) {
groupingKeywords.push(match[1].trim())
}
return groupingKeywords
}
/**
* Suggest a query based on natural language input
*/
export function suggestQuery(
metadata: CubeMetadata[],
naturalLanguage: string,
targetCube?: string
): QuerySuggestion {
const reasoning: string[] = []
const warnings: string[] = []
const query: Partial<SemanticQuery> = {}
// Detect analysis mode
const analysisMode = detectAnalysisMode(naturalLanguage)
// Step 1: Discover relevant cubes if not specified
let relevantCubes: CubeMetadata[]
if (targetCube) {
const cube = metadata.find(c => c.name === targetCube)
if (cube) {
relevantCubes = [cube]
reasoning.push(`Using specified cube: ${targetCube}`)
} else {
warnings.push(`Specified cube '${targetCube}' not found`)
relevantCubes = []
}
} else {
const discoveryResults = discoverCubes(metadata, { intent: naturalLanguage, limit: 3 })
relevantCubes = discoveryResults
.map(r => metadata.find(c => c.name === r.cube))
.filter((c): c is CubeMetadata => c !== undefined)
if (relevantCubes.length > 0) {
reasoning.push(`Identified relevant cubes: ${relevantCubes.map(c => c.name).join(', ')}`)
}
}
if (relevantCubes.length === 0) {
// For analysis modes, still provide nextSteps guidance even without cubes
// Return confidence 0.7 for analysis modes (mode was detected), 0 for query mode
const isAnalysisMode = analysisMode !== 'query'
const nextSteps = isAnalysisMode
? generateNextSteps(analysisMode, undefined)
: undefined
return {
query: {},
confidence: isAnalysisMode ? 0.7 : 0,
reasoning: isAnalysisMode
? [`Detected ${analysisMode} intent from natural language`]
: ['Could not identify relevant cubes for this query'],
warnings,
analysisMode,
nextSteps
}
}
const primaryCube = relevantCubes[0]
let confidence = 0.5 // Base confidence
// Step 2: Detect aggregation intent
const aggregationIntent = detectAggregationIntent(naturalLanguage)
if (aggregationIntent) {
reasoning.push(`Detected ${aggregationIntent.type} aggregation intent`)
confidence += 0.1
}
// Step 3: Find matching measures
const measures: string[] = []
const lowerText = naturalLanguage.toLowerCase()
// Look for measure keywords in the text
for (const measure of primaryCube.measures) {
const measureName = measure.name.split('.').pop() || measure.name
// Check name, title, and synonyms
const namesToCheck = [
measureName.toLowerCase(),
measure.title.toLowerCase(),
...(measure.synonyms || []).map(s => s.toLowerCase())
]
for (const name of namesToCheck) {
if (lowerText.includes(name)) {
measures.push(measure.name)
reasoning.push(`Matched measure '${measure.name}' via keyword '${name}'`)
confidence += 0.15
break
}
}
}
// If no specific measures found, suggest based on aggregation intent
if (measures.length === 0 && aggregationIntent) {
// Find measures matching the aggregation type
const matchingMeasures = primaryCube.measures.filter(m => m.type === aggregationIntent.type)
if (matchingMeasures.length > 0) {
measures.push(matchingMeasures[0].name)
reasoning.push(`Suggested ${matchingMeasures[0].name} based on ${aggregationIntent.type} intent`)
E} else if (aggregationIntent.type === 'count') {
// For count, find any count or countDistinct measure
const countMeasure = primaryCube.measures.find(m =>
m.type === 'count' || m.type === 'countDistinct'
)
if (countMeasure) {
measures.push(countMeasure.name)
reasoning.push(`Suggested ${countMeasure.name} for counting`)
}
}
}
// If still no measures, use first available
if (measures.length === 0 && primaryCube.measures.length > 0) {
measures.push(primaryCube.measures[0].name)
reasoning.push(`Using default measure: ${primaryCube.measures[0].name}`)
warnings.push('Could not determine specific measure from query, using default')
}
query.measures = measures
// Step 4: Detect and resolve grouping dimensions
const groupingKeywords = detectGroupingIntent(naturalLanguage)
const dimensions: string[] = []
for (const keyword of groupingKeywords) {
const match = findBestFieldMatch(relevantCubes, keyword, 'dimension')
if (match) {
dimensions.push(match.field)
reasoning.push(`Matched dimension '${match.field}' from grouping keyword '${keyword}'`)
confidence += 0.1
}
}
// Also check for dimension keywords in the text
for (const cube of relevantCubes) {
for (const dimension of cube.dimensions) {
const dimName = dimension.name.split('.').pop() || dimension.name
const namesToCheck = [
dimName.toLowerCase(),
dimension.title.toLowerCase(),
...(dimension.synonyms || []).map(s => s.toLowerCase())
]
for (const name of namesToCheck) {
if (lowerText.includes(name) && !dimensions.includes(dimension.name)) {
// Check if this is likely a grouping dimension
if (lowerText.includes(`by ${name}`) || lowerText.includes(`per ${name}`)) {
dimensions.push(dimension.name)
reasoning.push(`Matched dimension '${dimension.name}' as grouping`)
confidence += 0.1
break
}
}
}
}
}
if (dimensions.length > 0) {
query.dimensions = dimensions
}
// Step 5: Parse time expressions
const timeExpr = parseTimeExpression(naturalLanguage)
if (timeExpr) {
// Find a time dimension in the primary cube
const timeDimension = primaryCube.dimensions.find(d => d.type === 'time')
if (timeDimension) {
const td: TimeDimension = {
dimension: timeDimension.name,
dateRange: timeExpr.dateRange
}
Eif (timeExpr.granularity) {
td.granularity = timeExpr.granularity as any
}
query.timeDimensions = [td]
reasoning.push(`Applied time filter: ${timeExpr.dateRange[0]} to ${timeExpr.dateRange[1]}`)
confidence += 0.15
} else {
warnings.push('Time expression found but no time dimension in cube')
}
}
// Normalize confidence to 0-1
confidence = Math.min(1, confidence)
// For analysis modes, return guidance instead of building query
if (analysisMode !== 'query') {
const primaryCubeName = relevantCubes.length > 0 ? relevantCubes[0].name : undefined
return {
query: {},
confidence: 0.7,
reasoning: [
`Detected ${analysisMode} intent from natural language`,
...(primaryCubeName ? [`Found relevant cube: ${primaryCubeName}`] : [])
],
warnings: warnings.length > 0 ? warnings : undefined,
analysisMode,
nextSteps: generateNextSteps(analysisMode, primaryCubeName)
}
}
return {
query,
confidence,
reasoning,
warnings: warnings.length > 0 ? warnings : undefined,
analysisMode: 'query'
}
}
|