@tellescope/utilities 1.254.1 → 1.255.0

package/src/ai_summary.ts

@@ -0,0 +1,313 @@
+// Shared, transport-agnostic AI summary code.
+//
+// This module holds the *pure* parts of the Form "AI profile summary" feature so they can be
+// reused by both the webapp SDK loader (data_sources.ts) and the backend DB loader
+// (api/modules/ai_summary_context.ts). It must stay free of any SDK/DB/React imports — only
+// plain record shapes, the types-models Enduser, and constants.
+
+import {
+  AISummaryConfiguration,
+  AISummaryDataSource,
+  AISummaryDataSourceConfig,
+  Enduser,
+} from "@tellescope/types-models"
+import {
+  DEFAULT_AI_SUMMARY_DATA_SOURCE_LIMIT,
+  MAX_AI_SUMMARY_INPUT_TOKENS,
+} from "@tellescope/constants"
+
+/* ---------------------------------- formatters ---------------------------------- */
+
+export const stripHtml = (s: string) => s
+  .replace(/<style[^>]*>[\s\S]*?<\/style>/gi, '')
+  .replace(/<script[^>]*>[\s\S]*?<\/script>/gi, '')
+  .replace(/<[^>]+>/g, ' ')
+  .replace(/&nbsp;/gi, ' ')
+  .replace(/\s+/g, ' ')
+  .trim()
+
+export const fmt = (d: any) => {
+  if (!d) return ''
+  try { return new Date(d).toISOString().slice(0, 16).replace('T', ' ') } catch { return '' }
+}
+export const fmtFromMS = (ms: any) => {
+  if (typeof ms !== 'number') return ''
+  try { return new Date(ms).toISOString().slice(0, 16).replace('T', ' ') } catch { return '' }
+}
+export const fmtResponses = (responses: any) => {
+  if (!Array.isArray(responses)) return ''
+  return responses
+    .map((r: any) => {
+      const ans = r?.answer
+      const v = ans?.value
+      if (v === undefined || v === null) return ''
+      if (typeof v === 'string' || typeof v === 'number' || typeof v === 'boolean') {
+        return `${r.fieldTitle || ''}: ${v}`
+      }
+      if (Array.isArray(v)) return `${r.fieldTitle || ''}: ${v.map((x: any) => x?.label ?? x).join(', ')}`
+      return ''
+    })
+    .filter(Boolean)
+    .join(' | ')
+}
+export const fmtObservation = (o: any) => {
+  const m = o.measurement
+  // numeric value + unit when present; fall back to qualitative result ("positive", etc.)
+  const reading =
+    m && typeof m.value === 'number'
+      ? `${m.value}${m.unit ? ` ${m.unit}` : ''}`
+      : (o.qualitativeResult ?? '')
+
+  const name = o.type || o.code || o.category || ''      // what was measured
+  const extras = [
+    o.refRange ? `ref ${o.refRange}` : '',
+    o.statusIndicator ?? '',                              // High / Low / Normal
+  ].filter(Boolean).join(', ')
+
+  const label = name ? `${name}: ` : ''
+  const tail = extras ? ` (${extras})` : ''
+  return `[Observation ${fmt(o.timestamp ?? o.recordedAt ?? o.createdAt)}] ${label}${reading}${tail}`.trim()
+}
+
+/* ---------------------------------- data source metadata ---------------------------------- */
+
+export const DATA_SOURCE_LABELS: Record<AISummaryDataSource, string> = {
+  enduser_observations: 'Observations',
+  form_responses: 'Form Responses',
+  chats: 'Chat Messages',
+  phone_calls: 'Phone Calls',
+  calendar_events: 'Calendar Events',
+  tickets: 'Tickets',
+  sms_messages: 'SMS Messages',
+  emails: 'Emails',
+  enduser_orders: 'Orders',
+  enduser_medications: 'Medications',
+  purchases: 'Purchases',
+}
+
+export type DataSourceMapEntry = {
+  // collection key equals the source key for all 11 entries; callers map this to their own
+  // transport (session.api[collection] / DB[collection]).
+  collection: AISummaryDataSource,
+  sortField: string,
+  format: (record: any) => string,
+  enduserMatchClause?: (enduserId: string) => object,
+}
+
+export const DATA_SOURCE_MAP: Record<AISummaryDataSource, DataSourceMapEntry> = {
+  enduser_observations: {
+    collection: 'enduser_observations',
+    sortField: 'timestamp',
+    format: fmtObservation,
+  },
+  form_responses: {
+    collection: 'form_responses',
+    sortField: 'submittedAt',
+    format: fr => `[Form "${fr.formTitle ?? ''}" ${fmt(fr.submittedAt ?? fr.createdAt)}] ${fmtResponses(fr.responses)}`,
+  },
+  chats: {
+    collection: 'chats',
+    sortField: 'timestamp',
+    // enduserId is set only on enduser-authored chats (schema initializer); senderId is always set
+    // (userId for staff, enduserId for patients) so it can't distinguish the sender.
+    format: m => `[Chat ${fmt(m.timestamp ?? m.createdAt)} ${m.enduserId ? 'patient' : 'staff'}] ${m.message ?? ''}`,
+  },
+  phone_calls: {
+    collection: 'phone_calls',
+    sortField: 'createdAt',
+    format: c => `[Call ${fmt(c.createdAt)} ${c.direction ?? ''} ${c.durationInSeconds ?? 0}s] ${c.aiSummary ?? c.summary ?? ''}`,
+  },
+  calendar_events: {
+    collection: 'calendar_events',
+    sortField: 'startTimeInMS',
+    format: e => `[Appt ${fmtFromMS(e.startTimeInMS)}] ${e.title ?? ''} (${e.type ?? ''})`,
+    enduserMatchClause: enduserId => ({ 'attendees.id': enduserId }),
+  },
+  tickets: {
+    collection: 'tickets',
+    sortField: 'createdAt',
+    format: t => `[Ticket ${fmt(t.createdAt)} ${t.closedAt ? 'closed' : 'open'}] ${t.title ?? ''}`,
+  },
+  sms_messages: {
+    collection: 'sms_messages',
+    sortField: 'timestamp',
+    format: s => `[SMS ${fmt(s.timestamp ?? s.createdAt)} ${s.inbound ? 'in' : 'out'}] ${s.message ?? ''}`,
+  },
+  emails: {
+    collection: 'emails',
+    sortField: 'timestamp',
+    format: e => {
+      const body = e.textContent || stripHtml(e.HTMLContent || '')
+      return `[Email ${fmt(e.timestamp ?? e.createdAt)}] ${e.subject ?? ''}\n${body.slice(0, 2000)}`
+    },
+  },
+  enduser_orders: {
+    collection: 'enduser_orders',
+    sortField: 'createdAt',
+    format: o => `[Order ${fmt(o.createdAt)}] ${o.status ?? ''} — ${(o.items ?? []).map((i: any) => i.title ?? i.sku ?? '').filter(Boolean).join(', ')}`,
+  },
+  enduser_medications: {
+    collection: 'enduser_medications',
+    sortField: 'createdAt',
+    format: m => `[Medication ${fmt(m.createdAt)}] ${m.title ?? ''} ${m.dosage ?? ''} ${m.status ?? ''}`.trim(),
+  },
+  purchases: {
+    collection: 'purchases',
+    sortField: 'createdAt',
+    format: p => `[Purchase ${fmt(p.createdAt)}] ${p.title ?? ''} ${typeof p.amount === 'number' ? `$${p.amount}` : ''}`.trim(),
+  },
+}
+
+export const enduserProfileToText = (e: Enduser): string => {
+  const lines: string[] = []
+  lines.push('## enduser')
+  if (e.fname || e.lname) lines.push(`Name: ${e.fname ?? ''} ${e.lname ?? ''}`.trim())
+  if (e.dateOfBirth) lines.push(`DOB: ${e.dateOfBirth}`)
+  if (e.gender) lines.push(`Gender: ${e.gender}`)
+  if (e.state) lines.push(`State: ${e.state}`)
+  if (e.email) lines.push(`Email: ${e.email}`)
+  if (e.phone) lines.push(`Phone: ${e.phone}`)
+  if (e.tags?.length) lines.push(`Tags: ${e.tags.join(', ')}`)
+  if (e.fields && typeof e.fields === 'object') {
+    const entries = Object.entries(e.fields).filter(([_, v]) => v !== '' && v !== null && v !== undefined)
+    if (entries.length) {
+      lines.push('Custom fields:')
+      for (const [k, v] of entries) {
+        lines.push(`  - ${k}: ${typeof v === 'object' ? JSON.stringify(v) : String(v)}`)
+      }
+    }
+  }
+  return lines.join('\n')
+}
+
+/* ---------------------------------- filter sanitization ---------------------------------- */
+
+export const FORBIDDEN_FILTER_OPERATORS = ['$where', '$function']
+export const aiSummaryError = (code: number, message: string) => Object.assign(new Error(message), { code })
+
+export const sanitizeFilter = (filter: any) => {
+  if (!filter || typeof filter !== 'object') return {}
+  const json = JSON.stringify(filter)
+  for (const op of FORBIDDEN_FILTER_OPERATORS) {
+    if (json.includes(op)) {
+      throw aiSummaryError(400, `AI summary filter cannot contain ${op}`)
+    }
+  }
+  return filter
+}
+
+/* ---------------------------------- context assembly ---------------------------------- */
+
+export type AISummarySourceSection = {
+  type: AISummaryDataSource,
+  label: string,
+  lookbackMS?: number,
+  limit: number,
+  records: any[],
+  formattedLines: string[],
+}
+
+export type AISummaryContext = {
+  contextText: string,
+  estimatedTokens: number,
+  profileBlock: string,
+  sources: AISummarySourceSection[],
+}
+
+// Build the mongo filter + effective limit for a single data source. Shared by both loaders so
+// the lookback / enduserMatch / sanitized-filter / default-limit logic stays identical.
+export const buildAISummarySourceFilter = ({ ds, enduserId, mapEntry } : {
+  ds: AISummaryDataSourceConfig,
+  enduserId: string,
+  mapEntry: DataSourceMapEntry,
+}): { mdbFilter: object, effectiveLimit: number } => {
+  const userFilter = sanitizeFilter(ds.filter)
+  const lookbackClause = ds.lookbackMS
+    ? { [mapEntry.sortField]: { $gte: new Date(Date.now() - ds.lookbackMS) } }
+    : {}
+
+  const enduserMatch = mapEntry.enduserMatchClause ? mapEntry.enduserMatchClause(enduserId) : { enduserId }
+  const mdbFilter = { $and: [userFilter, lookbackClause, enduserMatch] }
+  const effectiveLimit = ds.limit ?? DEFAULT_AI_SUMMARY_DATA_SOURCE_LIMIT
+
+  return { mdbFilter, effectiveLimit }
+}
+
+// Join the profile block + per-source blocks, estimate tokens, and enforce the input budget.
+export const assembleAISummaryContext = ({ profileBlock, sources } : {
+  profileBlock: string,
+  sources: AISummarySourceSection[],
+}): AISummaryContext => {
+  const blocks = sources
+    .filter(s => s.formattedLines.length > 0)
+    .map(s => `## ${s.type}\n${s.formattedLines.join('\n')}`)
+
+  const contextText = [profileBlock, ...blocks].filter(Boolean).join('\n\n')
+  const estimatedTokens = Math.ceil(contextText.length / 4)
+  if (estimatedTokens > MAX_AI_SUMMARY_INPUT_TOKENS) {
+    throw aiSummaryError(413, 'AI summary input exceeds token budget — reduce limits, shorten lookback, or remove data sources.')
+  }
+  return { contextText, estimatedTokens, profileBlock, sources }
+}
+
+/* ---------------------------------- transport-agnostic loader ---------------------------------- */
+
+export type LoadAISummaryRecordsArgs = {
+  type: AISummaryDataSource,
+  collection: AISummaryDataSource,
+  mdbFilter: object,
+  limit: number,
+  sortField: string,
+}
+
+// Transport-agnostic loader. Callers inject loadProfile/loadRecords backed by the SDK (webapp) or
+// the DB (api). Per-source failures degrade to an empty section so one broken collection doesn't
+// sink the whole summary; filter-sanitization (400) and the token-budget (413) errors still throw.
+export const loadAISummaryContext = async ({
+  enduserId, configuration, loadProfile, loadRecords, includeProfile = true,
+} : {
+  enduserId: string,
+  configuration: AISummaryConfiguration,
+  loadProfile: (enduserId: string) => Promise<Enduser>,
+  loadRecords: (args: LoadAISummaryRecordsArgs) => Promise<any[]>,
+  // include the enduser profile block (name/DOB/fields/custom fields) in the context. Defaults to
+  // true for the summary use case; the AI Decision step passes false to avoid prompt clutter.
+  includeProfile?: boolean,
+}): Promise<AISummaryContext> => {
+  const enduser = includeProfile ? await loadProfile(enduserId) : null
+  const profileBlock = enduser ? enduserProfileToText(enduser) : ''
+
+  const sections = await Promise.all((configuration.dataSources ?? []).map(async (ds: AISummaryDataSourceConfig): Promise<AISummarySourceSection | null> => {
+    const m = DATA_SOURCE_MAP[ds.type]
+    if (!m) return null
+
+    const { mdbFilter, effectiveLimit } = buildAISummarySourceFilter({ ds, enduserId, mapEntry: m })
+
+    try {
+      const records = await loadRecords({ type: ds.type, collection: m.collection, mdbFilter, limit: effectiveLimit, sortField: m.sortField })
+      const list: any[] = Array.isArray(records) ? records : []
+      const formattedLines = list.map(m.format).filter(Boolean)
+      return {
+        type: ds.type,
+        label: DATA_SOURCE_LABELS[ds.type] ?? ds.type,
+        lookbackMS: ds.lookbackMS,
+        limit: effectiveLimit,
+        records: list,
+        formattedLines,
+      }
+    } catch (err) {
+      console.error(`Failed to load ${ds.type} for AI summary`, err)
+      return {
+        type: ds.type,
+        label: DATA_SOURCE_LABELS[ds.type] ?? ds.type,
+        lookbackMS: ds.lookbackMS,
+        limit: effectiveLimit,
+        records: [],
+        formattedLines: [],
+      }
+    }
+  }))
+
+  const sources = sections.filter((s): s is AISummarySourceSection => s !== null)
+  return assembleAISummaryContext({ profileBlock, sources })
+}

package/src/utils.ts

@@ -6,6 +6,8 @@ import { DateTime } from "luxon"
 import { ObjectId } from "./ObjectId/objectid";
 export { ObjectId }
 
+export * from "./ai_summary"
+
 export type Indexable<T=any> = { [index: string]: T }
 
 export const user_is_admin = (u: { id: string, roles?: string[] } & ({ type: 'user' } | { type: 'enduser' })) =>