#!/usr/bin/env npx tsx
/**
 * MCP Speech Transcription Tool — Monetized with SettleGrid
 *
 * A complete MCP server that transcribes audio via OpenAI Whisper and
 * summarizes transcripts via Claude. Fork, add your keys, and deploy.
 *
 * Setup:
 *   1. npm install @settlegrid/mcp
 *   2. Set OPENAI_API_KEY, ANTHROPIC_API_KEY, and SETTLEGRID_API_KEY in your env
 *   3. Register your tool at settlegrid.ai/dashboard/tools
 *   4. Run: npx tsx mcp-speech-transcription.ts
 *
 * Pricing: 5 cents/min transcription, 8 cents/min with speakers, 4 cents per summary
 *   - Whisper API costs $0.006/min
 *   - 5 cents = ~8x margin on basic transcription
 *   - Speaker diarization adds post-processing, 8 cents = ~5x margin
 *   - Claude summary costs ~$0.008, 4 cents = ~5x margin
 *
 * Revenue: You keep 95-100% (100% on Free tier, 95% on paid tiers)
 */

import { settlegrid } from '@settlegrid/mcp'

// ── SettleGrid Setup ────────────────────────────────────────────────────────

const sg = settlegrid.init({
  toolSlug: 'my-speech-transcription', // Replace with your tool slug
  pricing: {
    defaultCostCents: 5,
    methods: {
      transcribe: { costCents: 5, displayName: 'Transcribe Audio' },
      transcribe_with_speakers: { costCents: 8, displayName: 'Transcribe with Speakers' },
      summarize_transcript: { costCents: 4, displayName: 'Summarize Transcript' },
    },
  },
})

// ── Whisper API Helper ──────────────────────────────────────────────────────

const AUDIO_URL_RE = /^https?:\/\/.+\.(mp3|mp4|wav|m4a|webm|ogg|flac|mpeg)(\?.*)?$/i

async function callWhisper(audioUrl: string, language?: string): Promise<Record<string, unknown>> {
  if (!audioUrl || !AUDIO_URL_RE.test(audioUrl)) {
    throw new Error('A valid audio URL is required (mp3, mp4, wav, m4a, webm, ogg, flac)')
  }
  const audioResponse = await fetch(audioUrl)
  if (!audioResponse.ok) throw new Error(`Failed to fetch audio: ${audioResponse.status}`)
  const audioBlob = await audioResponse.blob()
  if (audioBlob.size > 25 * 1024 * 1024) throw new Error('Audio file exceeds 25 MB limit')

  const form = new FormData()
  form.append('file', audioBlob, 'audio.mp3')
  form.append('model', 'whisper-1')
  form.append('response_format', 'verbose_json')
  if (language) form.append('language', language)

  const response = await fetch('https://api.openai.com/v1/audio/transcriptions', {
    method: 'POST',
    headers: { Authorization: `Bearer ${process.env.OPENAI_API_KEY!}` },
    body: form,
  })
  if (!response.ok) throw new Error(`Whisper API returned ${response.status}: ${response.statusText}`)
  return response.json()
}

// ── Transcription Methods ───────────────────────────────────────────────────

interface TranscribeArgs { audioUrl: string; language?: string }
interface Segment { start: number; end: number; text: string }

async function transcribe(args: TranscribeArgs): Promise<{
  result: { text: string; language: string; duration: number; segments: Segment[] }
}> {
  const data = await callWhisper(args.audioUrl, args.language)
  const segments: Segment[] = ((data.segments as Array<Record<string, unknown>>) ?? []).map((s) => ({
    start: s.start as number, end: s.end as number, text: (s.text as string).trim(),
  }))
  return { result: { text: (data.text as string) ?? '', language: (data.language as string) ?? 'unknown', duration: (data.duration as number) ?? 0, segments } }
}

async function transcribeWithSpeakers(args: TranscribeArgs): Promise<{
  result: { text: string; language: string; duration: number; segments: Array<Segment & { speaker: string }>; speakerCount: number }
}> {
  const data = await callWhisper(args.audioUrl, args.language)
  const rawSegments = (data.segments as Array<Record<string, unknown>>) ?? []
  let currentSpeaker = 0
  const segments = rawSegments.map((s, i) => {
    if (i > 0) {
      const gap = (s.start as number) - (rawSegments[i - 1].end as number)
      if (gap > 2) currentSpeaker = (currentSpeaker + 1) % 10
    }
    return { start: s.start as number, end: s.end as number, text: (s.text as string).trim(), speaker: `Speaker ${currentSpeaker + 1}` }
  })
  return { result: { text: (data.text as string) ?? '', language: (data.language as string) ?? 'unknown', duration: (data.duration as number) ?? 0, segments, speakerCount: new Set(segments.map((s) => s.speaker)).size } }
}

interface SummarizeArgs { transcript: string; style?: 'bullets' | 'paragraph' | 'action-items' }

async function summarizeTranscript(args: SummarizeArgs): Promise<{ summary: string }> {
  if (!args.transcript || args.transcript.trim().length === 0) throw new Error('Transcript text must be non-empty')
  if (args.transcript.length > 100_000) throw new Error('Transcript exceeds 100,000 character limit')
  const style = args.style ?? 'bullets'
  const response = await fetch('https://api.anthropic.com/v1/messages', {
    method: 'POST',
    headers: { 'Content-Type': 'application/json', 'x-api-key': process.env.ANTHROPIC_API_KEY!, 'anthropic-version': '2023-06-01' },
    body: JSON.stringify({
      model: 'claude-sonnet-4-20250514', max_tokens: 2048,
      system: `You are a transcript summarizer. Produce a ${style} summary covering key topics, decisions, and action items.`,
      messages: [{ role: 'user', content: `Summarize this transcript:\n\n${args.transcript}` }],
    }),
  })
  if (!response.ok) throw new Error(`Claude API returned ${response.status}: ${response.statusText}`)
  const data = await response.json()
  return { summary: data.content?.[0]?.text ?? '' }
}

// ── Wrap with SettleGrid Billing ─────────────────────────────────────────────

export const billedTranscribe = sg.wrap(transcribe, { method: 'transcribe' })
export const billedTranscribeWithSpeakers = sg.wrap(transcribeWithSpeakers, { method: 'transcribe_with_speakers' })
export const billedSummarize = sg.wrap(summarizeTranscript, { method: 'summarize_transcript' })

// ── REST Alternative ────────────────────────────────────────────────────────
// import { settlegridMiddleware } from '@settlegrid/mcp/rest'
//
// const withBilling = settlegridMiddleware({
//   toolSlug: 'my-speech-transcription',
//   pricing: {
//     defaultCostCents: 5,
//     methods: {
//       transcribe: { costCents: 5 },
//       transcribe_with_speakers: { costCents: 8 },
//       summarize_transcript: { costCents: 4 },
//     },
//   },
// })
//
// export async function POST(request: Request) {
//   return withBilling(request, async () => {
//     const { audioUrl } = await request.json()
//     const result = await transcribe({ audioUrl })
//     return Response.json(result)
//   }, 'transcribe')
// }