nyuchi-fundi-learning

lib

Fundi learning system. Tracks healing outcomes to improve future decisions. When an issue is resolved, records what failed, what Fundi suggested, what actually worked, and whether the same pattern recurs. Runs as part of the Fundi Cloudflare Worker — same Rust binary, different handler. Writes learned patterns to fundi_issues via Supabase REST API. Improves classification accuracy over time.

Source code

No interactive demo is registered for this component yet — the source is shown directly.

"use client"

export interface HealingOutcome {
  issueId: number
  component: string
  layer: number
  errorType: string
  severity: string
  planActions: string[]
  actualFix: string
  fundiWasCorrect: boolean
  timeToResolveMinutes: number
  recurred: boolean
  resolvedBy: "fundi" | "human" | "both"
  timestamp: string
}

export interface LearningStats {
  totalIssues: number
  autoFixed: number
  humanFixed: number
  accuracy: number
  avgTimeToResolve: number
  topFailingComponents: { name: string; count: number }[]
  topErrorTypes: { type: string; count: number }[]
  recurrenceRate: number
}

// DB PERSISTENCE: In production, outcomes are stored in fundi_healing_log table
// and stats are computed from fundi_issues + fundi_healing_log via SQL.
// This in-memory implementation is for client-side quick access;
// the edge function (fundi-webhook) handles the DB persistence server-side.
//
// To query learning stats from DB:
// SELECT COUNT(*) as total, COUNT(*) FILTER (WHERE resolved_by = 'fundi') as auto_fixed,
//   AVG(EXTRACT(EPOCH FROM resolved_at - created_at)/60) as avg_ttr_minutes
// FROM fundi_issues WHERE status = 'resolved';

class FundiLearningCore {
  private outcomes: HealingOutcome[] = []

  record(outcome: HealingOutcome) {
    this.outcomes.push(outcome)
  }

  getStats(): LearningStats {
    const total = this.outcomes.length
    if (total === 0) return { totalIssues: 0, autoFixed: 0, humanFixed: 0, accuracy: 0, avgTimeToResolve: 0, topFailingComponents: [], topErrorTypes: [], recurrenceRate: 0 }

    const autoFixed = this.outcomes.filter(o => o.resolvedBy === "fundi").length
    const correct = this.outcomes.filter(o => o.fundiWasCorrect).length
    const avgTTR = this.outcomes.reduce((s, o) => s + o.timeToResolveMinutes, 0) / total
    const recurred = this.outcomes.filter(o => o.recurred).length

    const compCounts = new Map<string, number>()
    const typeCounts = new Map<string, number>()
    this.outcomes.forEach(o => {
      compCounts.set(o.component, (compCounts.get(o.component) || 0) + 1)
      typeCounts.set(o.errorType, (typeCounts.get(o.errorType) || 0) + 1)
    })

    return {
      totalIssues: total,
      autoFixed,
      humanFixed: total - autoFixed,
      accuracy: correct / total,
      avgTimeToResolve: Math.round(avgTTR),
      topFailingComponents: [...compCounts.entries()].sort((a, b) => b[1] - a[1]).slice(0, 10).map(([name, count]) => ({ name, count })),
      topErrorTypes: [...typeCounts.entries()].sort((a, b) => b[1] - a[1]).slice(0, 5).map(([type, count]) => ({ type, count })),
      recurrenceRate: recurred / total,
    }
  }

  getComponentHistory(component: string): HealingOutcome[] {
    return this.outcomes.filter(o => o.component === component)
  }

  suggestSeverity(component: string, errorType: string): string {
    const history = this.outcomes.filter(o => o.component === component && o.errorType === errorType)
    if (history.length === 0) return "medium"
    const lastSeverity = history[history.length - 1].severity
    const recurrenceCount = history.filter(o => o.recurred).length
    if (recurrenceCount > 2) return "high"
    return lastSeverity
  }
}

let _learning: FundiLearningCore | null = null
export function getFundiLearning(): FundiLearningCore {
  if (!_learning) _learning = new FundiLearningCore()
  return _learning
}
export type { FundiLearningCore }

Variants & props

The variants, sizes, and props surfaced by the registry. Each is sourced from the Supabase component_docs table.

API tester

Hit the registry API live and see the JSON the shadcn CLI consumes when it installs this component.

GET/api/v1/ui/nyuchi-fundi-learning

Install

npx shadcn@latest add https://mzizi.dev/api/v1/ui/nyuchi-fundi-learning

Source path

lib/nyuchi-fundi-learning.ts

Source code

No interactive demo is registered for this component yet — the source is shown directly.

"use client"

export interface HealingOutcome {
  issueId: number
  component: string
  layer: number
  errorType: string
  severity: string
  planActions: string[]
  actualFix: string
  fundiWasCorrect: boolean
  timeToResolveMinutes: number
  recurred: boolean
  resolvedBy: "fundi" | "human" | "both"
  timestamp: string
}

export interface LearningStats {
  totalIssues: number
  autoFixed: number
  humanFixed: number
  accuracy: number
  avgTimeToResolve: number
  topFailingComponents: { name: string; count: number }[]
  topErrorTypes: { type: string; count: number }[]
  recurrenceRate: number
}

// DB PERSISTENCE: In production, outcomes are stored in fundi_healing_log table
// and stats are computed from fundi_issues + fundi_healing_log via SQL.
// This in-memory implementation is for client-side quick access;
// the edge function (fundi-webhook) handles the DB persistence server-side.
//
// To query learning stats from DB:
// SELECT COUNT(*) as total, COUNT(*) FILTER (WHERE resolved_by = 'fundi') as auto_fixed,
//   AVG(EXTRACT(EPOCH FROM resolved_at - created_at)/60) as avg_ttr_minutes
// FROM fundi_issues WHERE status = 'resolved';

class FundiLearningCore {
  private outcomes: HealingOutcome[] = []

  record(outcome: HealingOutcome) {
    this.outcomes.push(outcome)
  }

  getStats(): LearningStats {
    const total = this.outcomes.length
    if (total === 0) return { totalIssues: 0, autoFixed: 0, humanFixed: 0, accuracy: 0, avgTimeToResolve: 0, topFailingComponents: [], topErrorTypes: [], recurrenceRate: 0 }

    const autoFixed = this.outcomes.filter(o => o.resolvedBy === "fundi").length
    const correct = this.outcomes.filter(o => o.fundiWasCorrect).length
    const avgTTR = this.outcomes.reduce((s, o) => s + o.timeToResolveMinutes, 0) / total
    const recurred = this.outcomes.filter(o => o.recurred).length

    const compCounts = new Map<string, number>()
    const typeCounts = new Map<string, number>()
    this.outcomes.forEach(o => {
      compCounts.set(o.component, (compCounts.get(o.component) || 0) + 1)
      typeCounts.set(o.errorType, (typeCounts.get(o.errorType) || 0) + 1)
    })

    return {
      totalIssues: total,
      autoFixed,
      humanFixed: total - autoFixed,
      accuracy: correct / total,
      avgTimeToResolve: Math.round(avgTTR),
      topFailingComponents: [...compCounts.entries()].sort((a, b) => b[1] - a[1]).slice(0, 10).map(([name, count]) => ({ name, count })),
      topErrorTypes: [...typeCounts.entries()].sort((a, b) => b[1] - a[1]).slice(0, 5).map(([type, count]) => ({ type, count })),
      recurrenceRate: recurred / total,
    }
  }

  getComponentHistory(component: string): HealingOutcome[] {
    return this.outcomes.filter(o => o.component === component)
  }

  suggestSeverity(component: string, errorType: string): string {
    const history = this.outcomes.filter(o => o.component === component && o.errorType === errorType)
    if (history.length === 0) return "medium"
    const lastSeverity = history[history.length - 1].severity
    const recurrenceCount = history.filter(o => o.recurred).length
    if (recurrenceCount > 2) return "high"
    return lastSeverity
  }
}

let _learning: FundiLearningCore | null = null
export function getFundiLearning(): FundiLearningCore {
  if (!_learning) _learning = new FundiLearningCore()
  return _learning
}
export type { FundiLearningCore }