gigafibre-fsm/services/targo-hub/lib/vision.js

'use strict'
const cfg = require('./config')
const { log, json, parseBody } = require('./helpers')

const GEMINI_URL = () => `https://generativelanguage.googleapis.com/v1beta/models/${cfg.AI_MODEL}:generateContent?key=${cfg.AI_API_KEY}`

async function geminiVision (base64Image, prompt, schema) {
  const resp = await fetch(GEMINI_URL(), {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({
      contents: [{ parts: [{ text: prompt }, { inline_data: { mime_type: 'image/jpeg', data: base64Image } }] }],
      generationConfig: { temperature: 0.1, maxOutputTokens: 1024, responseMimeType: 'application/json', responseSchema: schema },
    }),
  })
  if (!resp.ok) { const t = await resp.text(); throw new Error(`Gemini API ${resp.status}: ${t.slice(0, 200)}`) }
  const data = await resp.json()
  const text = (data.candidates?.[0]?.content?.parts?.[0]?.text || '').trim()
  log(`Vision response: ${text.slice(0, 300)}`)
  let parsed
  try { parsed = JSON.parse(text) } catch { const m = text.match(/\{[\s\S]*\}/); if (m) try { parsed = JSON.parse(m[0]) } catch {} }
  return parsed
}

function extractBase64 (req, body, label) {
  if (!cfg.AI_API_KEY) return { error: 'AI_API_KEY not configured', status: 500 }
  if (!body.image) return { error: 'Missing image field (base64)', status: 400 }
  const base64 = body.image.replace(/^data:image\/[^;]+;base64,/, '')
  log(`Vision ${label}: received image ${Math.round(base64.length * 3 / 4 / 1024)}KB`)
  return { base64 }
}

const BARCODE_PROMPT = `Read ALL identifiers on this equipment label photo (may be blurry/tilted).
Extract: barcode text, serial numbers (S/N, SN), MAC addresses (12 hex chars), model numbers (M/N, Model, P/N), IMEI, GPON SN.
Examples: 1608K44D9E79FAFF5, TPLG-A1B2C3D4, 04:18:D6:A1:B2:C3, HWTC87654321.
Try your BEST on every character. Return max 3 most important (serial/MAC first).`

const BARCODE_SCHEMA = {
  type: 'object',
  properties: { barcodes: { type: 'array', items: { type: 'string' }, maxItems: 3 } },
  required: ['barcodes'],
}

async function handleBarcodes (req, res) {
  const body = await parseBody(req)
  const check = extractBase64(req, body, 'barcode')
  if (check.error) return json(res, check.status, { error: check.error })
  try {
    const result = await extractBarcodes(check.base64)
    return json(res, 200, result)
  } catch (e) {
    log('Vision barcode error:', e.message)
    return json(res, 500, { error: 'Vision extraction failed: ' + e.message })
  }
}

async function extractBarcodes (base64Image) {
  const parsed = await geminiVision(base64Image, BARCODE_PROMPT, BARCODE_SCHEMA)
  if (!parsed) return { barcodes: [] }
  const arr = Array.isArray(parsed) ? parsed : Array.isArray(parsed.barcodes) ? parsed.barcodes : []
  const barcodes = arr.filter(v => typeof v === 'string' && v.trim().length > 3).map(v => v.trim().replace(/\s+/g, '')).slice(0, 3)
  log(`Vision: extracted ${barcodes.length} barcode(s): ${barcodes.join(', ')}`)
  return { barcodes }
}

const EQUIP_PROMPT = `Read this ISP equipment label (ONT/ONU/router/modem). Return structured JSON.
Extract: brand/manufacturer, model (M/N, P/N), serial (S/N, SN, under barcode), MAC address (12 hex, no separators), GPON SN, HW version, barcodes.
Try your BEST on blurry/angled text. Set missing fields to null.`

const EQUIP_SCHEMA = {
  type: 'object',
  properties: {
    brand: { type: 'string', nullable: true }, model: { type: 'string', nullable: true },
    serial_number: { type: 'string', nullable: true }, mac_address: { type: 'string', nullable: true },
    gpon_sn: { type: 'string', nullable: true }, hw_version: { type: 'string', nullable: true },
    equipment_type: { type: 'string', nullable: true },
    barcodes: { type: 'array', items: { type: 'string' }, maxItems: 5 },
  },
  required: ['serial_number'],
}

async function handleEquipment (req, res) {
  const body = await parseBody(req)
  const check = extractBase64(req, body, 'equipment')
  if (check.error) return json(res, check.status, { error: check.error })
  try {
    const parsed = await geminiVision(check.base64, EQUIP_PROMPT, EQUIP_SCHEMA)
    if (!parsed) return json(res, 200, { serial_number: null, barcodes: [] })
    if (parsed.mac_address) parsed.mac_address = parsed.mac_address.replace(/[:\-.\s]/g, '').toUpperCase()
    if (parsed.serial_number) parsed.serial_number = parsed.serial_number.replace(/\s+/g, '').trim()
    log(`Vision equipment: brand=${parsed.brand} model=${parsed.model} sn=${parsed.serial_number} mac=${parsed.mac_address}`)
    return json(res, 200, parsed)
  } catch (e) {
    log('Vision equipment error:', e.message)
    return json(res, 500, { error: 'Vision extraction failed: ' + e.message })
  }
}

// ─── Invoice / bill OCR ────────────────────────────────────────────────
// We run this on Gemini (not on Ollama) because the ops VM has no GPU —
// ops must not depend on a local vision model. The schema matches what
// the ops InvoiceScanPage expects so switching away from Ollama is a
// drop-in replacement on the frontend.

const INVOICE_PROMPT = `You are an invoice/bill OCR assistant. Extract structured data from this photo of a vendor invoice or bill.
Return ONLY valid JSON that matches the provided schema. No prose, no markdown.
Rules:
- "date" / "due_date" must be ISO YYYY-MM-DD. If the date is MM/DD/YYYY or DD/MM/YYYY and ambiguous, prefer YYYY-MM-DD with the most likely interpretation for Canadian/Québec invoices.
- "currency" is a 3-letter code (CAD, USD, EUR). Default to CAD if not visible.
- "tax_gst" = GST/TPS/HST (Canadian federal tax); "tax_qst" = QST/TVQ (Québec provincial tax).
- "items" is a line-by-line list; keep description as printed, collapse whitespace.
- Missing fields → null for strings, 0 for numbers, [] for items.`

const INVOICE_SCHEMA = {
  type: 'object',
  properties: {
    vendor:         { type: 'string', nullable: true },
    vendor_address: { type: 'string', nullable: true },
    invoice_number: { type: 'string', nullable: true },
    date:           { type: 'string', nullable: true },
    due_date:       { type: 'string', nullable: true },
    subtotal:       { type: 'number', nullable: true },
    tax_gst:        { type: 'number', nullable: true },
    tax_qst:        { type: 'number', nullable: true },
    total:          { type: 'number', nullable: true },
    currency:       { type: 'string', nullable: true },
    items: {
      type: 'array',
      items: {
        type: 'object',
        properties: {
          description: { type: 'string', nullable: true },
          qty:         { type: 'number', nullable: true },
          rate:        { type: 'number', nullable: true },
          amount:      { type: 'number', nullable: true },
        },
      },
    },
    notes: { type: 'string', nullable: true },
  },
  required: ['vendor', 'total'],
}

async function handleInvoice (req, res) {
  const body = await parseBody(req)
  const check = extractBase64(req, body, 'invoice')
  if (check.error) return json(res, check.status, { error: check.error })
  try {
    const parsed = await geminiVision(check.base64, INVOICE_PROMPT, INVOICE_SCHEMA)
    if (!parsed) return json(res, 200, { vendor: null, total: null, items: [] })
    // Normalize: trim + coerce numbers (model sometimes returns "1,234.56" as string)
    for (const k of ['subtotal', 'tax_gst', 'tax_qst', 'total']) {
      if (typeof parsed[k] === 'string') parsed[k] = Number(parsed[k].replace(/[^0-9.\-]/g, '')) || 0
    }
    if (Array.isArray(parsed.items)) {
      for (const it of parsed.items) {
        for (const k of ['qty', 'rate', 'amount']) {
          if (typeof it[k] === 'string') it[k] = Number(it[k].replace(/[^0-9.\-]/g, '')) || 0
        }
      }
    }
    log(`Vision invoice: vendor=${parsed.vendor} total=${parsed.total} items=${(parsed.items || []).length}`)
    return json(res, 200, parsed)
  } catch (e) {
    log('Vision invoice error:', e.message)
    return json(res, 500, { error: 'Vision extraction failed: ' + e.message })
  }
}

module.exports = { handleBarcodes, extractBarcodes, handleEquipment, handleInvoice }