agenciapsilmno/supabase/functions/whatsapp-heartbeat-check/index.ts

/*
|--------------------------------------------------------------------------
| Agência PSI — Edge Function: whatsapp-heartbeat-check
|--------------------------------------------------------------------------
| Cron a cada 2 minutos. Checa conexão Evolution de todos os tenants com
| WhatsApp Pessoal ativo. Atualiza connection_status + last_health_check.
|
| Fluxo por channel:
|   1. GET {api_url}/instance/connectionState/{instance_name}
|   2. Se state === 'open'  → marca connected, limpa first_unhealthy_at,
|                              resolve incidents abertos
|   3. Se state !== 'open' → mapeia pra connection_status, seta
|                             first_unhealthy_at (se não tinha), e:
|        - Se já passou `heartbeat_threshold_minutes` (default 5) desde
|          first_unhealthy_at → abre incident (idempotente) + notifica
|          admins ativos do tenant
|        - Senão só atualiza status (ainda não alerta)
|   4. Se erro HTTP / timeout → kind='error', segue mesma regra do caso 3
|
| Config por tenant (em notification_channels.metadata):
|   - heartbeat_threshold_minutes (default 5)
|   - heartbeat_alerts_enabled     (default true)
|
| Não alerta duas vezes pelo mesmo incident (checa notified_at).
|--------------------------------------------------------------------------
*/

import { createClient, SupabaseClient } from 'https://esm.sh/@supabase/supabase-js@2'

const corsHeaders = {
    'Access-Control-Allow-Origin':  '*',
    'Access-Control-Allow-Headers': 'authorization, x-client-info, apikey, content-type',
    'Access-Control-Allow-Methods': 'POST, GET, OPTIONS',
}

const DEFAULT_THRESHOLD_MINUTES = 5
const FETCH_TIMEOUT_MS = 8000
// Cooldown entre tentativas de reconnect por canal pra não martelar a Evolution
const RECONNECT_COOLDOWN_MS = 10 * 60 * 1000
// Quanto esperar depois de pedir restart antes de re-verificar o state
const RECONNECT_SETTLE_MS = 3000

function json(body: unknown, status = 200) {
    return new Response(JSON.stringify(body), {
        status,
        headers: { ...corsHeaders, 'Content-Type': 'application/json' }
    })
}

function rewriteForContainer(apiUrl: string): string {
    try {
        const u = new URL(apiUrl)
        if (u.hostname === 'localhost' || u.hostname === '127.0.0.1') {
            u.hostname = 'host.docker.internal'
            return u.toString().replace(/\/+$/, '')
        }
        return apiUrl.replace(/\/+$/, '')
    } catch {
        return apiUrl
    }
}

function mapStateToStatus(state: string | null): { status: string, kind: string } {
    switch (state) {
        case 'open':
            return { status: 'connected', kind: 'connected' }
        case 'connecting':
            return { status: 'connecting', kind: 'connecting' }
        case 'qr':
        case 'qrcode':
            return { status: 'qr_pending', kind: 'qr_pending' }
        case 'close':
        case 'closed':
            return { status: 'disconnected', kind: 'disconnected' }
        default:
            return { status: 'error', kind: 'unknown' }
    }
}

async function fetchWithTimeout(url: string, init: RequestInit, timeoutMs: number): Promise<Response> {
    const controller = new AbortController()
    const timer = setTimeout(() => controller.abort(), timeoutMs)
    try {
        return await fetch(url, { ...init, signal: controller.signal })
    } finally {
        clearTimeout(timer)
    }
}

interface ChannelRow {
    id: string
    tenant_id: string
    owner_id: string
    provider: string
    credentials: Record<string, string>
    connection_status: string | null
    last_health_check: string | null
    metadata: Record<string, unknown> | null
}

async function checkOneChannel(supa: SupabaseClient, channel: ChannelRow, now: Date): Promise<{
    tenant_id: string
    channel_id: string
    previous_status: string | null
    new_status: string
    action: 'ok' | 'opened' | 'resolved' | 'still_unhealthy' | 'no_change' | 'config_missing' | 'fetch_error' | 'auto_reconnected'
    incident_id?: string
    reconnect_attempted?: boolean
}> {
    const creds = channel.credentials || {}
    const apiUrl = String(creds.api_url || '').trim()
    const apiKey = String(creds.api_key || '').trim()
    const instance = String(creds.instance_name || '').trim()

    if (!apiUrl || !apiKey || !instance) {
        // Credencial incompleta — não alertamos, só marca error e segue
        await supa.from('notification_channels')
            .update({ connection_status: 'error', last_health_check: now.toISOString() })
            .eq('id', channel.id)
        return { tenant_id: channel.tenant_id, channel_id: channel.id, previous_status: channel.connection_status, new_status: 'error', action: 'config_missing' }
    }

    const base = rewriteForContainer(apiUrl)
    const targetUrl = `${base}/instance/connectionState/${encodeURIComponent(instance)}`

    let state: string | null = null
    let rawBody: unknown = null
    let fetchError: string | null = null
    try {
        const res = await fetchWithTimeout(targetUrl, {
            method: 'GET',
            headers: { 'Content-Type': 'application/json', apikey: apiKey }
        }, FETCH_TIMEOUT_MS)
        if (!res.ok) {
            fetchError = `http_${res.status}`
        } else {
            rawBody = await res.json().catch(() => null)
            // Evolution retorna: { instance: { instanceName, state }} OU { state }
            const body = rawBody as { instance?: { state?: string }, state?: string } | null
            state = body?.instance?.state ?? body?.state ?? null
        }
    } catch (e) {
        fetchError = (e as Error).message || 'fetch_failed'
    }

    const { status: newStatus, kind } = fetchError ? { status: 'error', kind: 'error' } : mapStateToStatus(state)
    const meta = (channel.metadata || {}) as Record<string, unknown>
    const thresholdMinutes = Number(meta.heartbeat_threshold_minutes) || DEFAULT_THRESHOLD_MINUTES
    const alertsEnabled = meta.heartbeat_alerts_enabled !== false
    const firstUnhealthyAtRaw = meta.first_unhealthy_at as string | undefined

    const patch: Record<string, unknown> = {
        connection_status: newStatus,
        last_health_check: now.toISOString()
    }
    const newMeta = { ...meta }

    if (newStatus === 'connected') {
        // Recuperou
        if (firstUnhealthyAtRaw) delete newMeta.first_unhealthy_at
        patch.metadata = newMeta

        await supa.from('notification_channels').update(patch).eq('id', channel.id)
        const { data: resolved } = await supa.rpc('whatsapp_heartbeat_resolve_open_incidents', { p_channel_id: channel.id })
        return {
            tenant_id: channel.tenant_id,
            channel_id: channel.id,
            previous_status: channel.connection_status,
            new_status: newStatus,
            action: (resolved as unknown as number) > 0 ? 'resolved' : 'ok'
        }
    }

    // Está unhealthy — seta first_unhealthy_at se ainda não tinha
    const firstUnhealthyAt = firstUnhealthyAtRaw ? new Date(firstUnhealthyAtRaw) : null
    if (!firstUnhealthyAt || isNaN(firstUnhealthyAt.getTime())) {
        newMeta.first_unhealthy_at = now.toISOString()
    }
    patch.metadata = newMeta
    await supa.from('notification_channels').update(patch).eq('id', channel.id)

    const minutesUnhealthy = firstUnhealthyAt ? (now.getTime() - firstUnhealthyAt.getTime()) / 60000 : 0

    if (minutesUnhealthy < thresholdMinutes) {
        return {
            tenant_id: channel.tenant_id,
            channel_id: channel.id,
            previous_status: channel.connection_status,
            new_status: newStatus,
            action: firstUnhealthyAt ? 'still_unhealthy' : 'no_change'
        }
    }

    // Passou do threshold — antes de abrir incident, tenta reconnect automático (6.3)
    // Só tenta se: reconnect habilitado + API respondeu (sem fetchError) + passou do cooldown
    const reconnectEnabled = meta.heartbeat_reconnect_enabled !== false
    const lastReconnectAtRaw = meta.heartbeat_reconnect_last_at as string | undefined
    const lastReconnectAt = lastReconnectAtRaw ? new Date(lastReconnectAtRaw) : null
    const canReconnect = reconnectEnabled
        && !fetchError
        && (!lastReconnectAt || isNaN(lastReconnectAt.getTime()) || (now.getTime() - lastReconnectAt.getTime()) >= RECONNECT_COOLDOWN_MS)

    let reconnectAttempted = false

    if (canReconnect) {
        reconnectAttempted = true
        try {
            // POST /instance/restart/{instance} — instrui Evolution a reconectar a sessão
            await fetchWithTimeout(`${base}/instance/restart/${encodeURIComponent(instance)}`, {
                method: 'POST',
                headers: { 'Content-Type': 'application/json', apikey: apiKey }
            }, FETCH_TIMEOUT_MS)

            // Aguarda estabilizar
            await new Promise((r) => setTimeout(r, RECONNECT_SETTLE_MS))

            // Rechecka o state
            const res2 = await fetchWithTimeout(targetUrl, {
                method: 'GET',
                headers: { 'Content-Type': 'application/json', apikey: apiKey }
            }, FETCH_TIMEOUT_MS)

            if (res2.ok) {
                const body2 = await res2.json().catch(() => null) as { instance?: { state?: string }, state?: string } | null
                const state2 = body2?.instance?.state ?? body2?.state ?? null

                if (state2 === 'open') {
                    // Reconnect funcionou: restaura estado como se nunca tivesse caído
                    const cleanedMeta = { ...newMeta }
                    delete cleanedMeta.first_unhealthy_at
                    cleanedMeta.heartbeat_reconnect_last_at = now.toISOString()
                    cleanedMeta.heartbeat_reconnect_count = (Number(cleanedMeta.heartbeat_reconnect_count) || 0) + 1

                    await supa.from('notification_channels').update({
                        connection_status: 'connected',
                        last_health_check: now.toISOString(),
                        metadata: cleanedMeta
                    }).eq('id', channel.id)

                    // Resolve qualquer incident aberto desse channel (caso tenha sobrado de ciclo anterior)
                    await supa.rpc('whatsapp_heartbeat_resolve_open_incidents', { p_channel_id: channel.id })

                    return {
                        tenant_id: channel.tenant_id,
                        channel_id: channel.id,
                        previous_status: channel.connection_status,
                        new_status: 'connected',
                        action: 'auto_reconnected',
                        reconnect_attempted: true
                    }
                }
            }
        } catch {
            // silencioso — reconnect falhou, segue fluxo de abrir incident
        }

        // Marca tentativa (mesmo que falhou) pra respeitar o cooldown
        newMeta.heartbeat_reconnect_last_at = now.toISOString()
        await supa.from('notification_channels').update({ metadata: newMeta }).eq('id', channel.id)
    }

    // Passou do threshold (e reconnect falhou / não tentou) — abre incident (idempotente)
    const incidentDetails = {
        ...(typeof rawBody === 'object' && rawBody ? rawBody : {}),
        ...(fetchError ? { error: fetchError } : {}),
        reconnect_attempted: reconnectAttempted
    }
    const { data: incidentId, error: incidentErr } = await supa.rpc('whatsapp_heartbeat_open_incident', {
        p_channel_id: channel.id,
        p_kind: kind,
        p_last_state: state || fetchError,
        p_details: incidentDetails
    })

    if (incidentErr) {
        return {
            tenant_id: channel.tenant_id,
            channel_id: channel.id,
            previous_status: channel.connection_status,
            new_status: newStatus,
            action: 'fetch_error'
        }
    }

    const newIncidentId = incidentId as unknown as string

    if (alertsEnabled && newIncidentId) {
        await notifyChannelStakeholders(supa, {
            tenant_id: channel.tenant_id,
            channel_owner_id: channel.owner_id,
            incident_id: newIncidentId,
            channel_display: String(channel.provider === 'evolution_api' ? 'WhatsApp Pessoal' : 'WhatsApp'),
            kind,
            minutes_unhealthy: Math.round(minutesUnhealthy)
        })
    }

    return {
        tenant_id: channel.tenant_id,
        channel_id: channel.id,
        previous_status: channel.connection_status,
        new_status: newStatus,
        action: 'opened',
        incident_id: newIncidentId,
        reconnect_attempted: reconnectAttempted
    }
}

async function notifyChannelStakeholders(supa: SupabaseClient, params: {
    tenant_id: string
    channel_owner_id: string
    incident_id: string
    channel_display: string
    kind: string
    minutes_unhealthy: number
}): Promise<void> {
    // Checa se já notificou esse incident
    const { data: incident } = await supa
        .from('whatsapp_connection_incidents')
        .select('notified_at, notification_count')
        .eq('id', params.incident_id)
        .maybeSingle()

    if (incident?.notified_at) return   // anti-spam: só notifica 1x pelo mesmo incident

    // Stakeholders = owner do canal + admins ativos do tenant (deduplicado).
    // owner geralmente é o dono do celular (WhatsApp Pessoal) ou admin da clínica;
    // admins garantem que alguém com permissão de infra seja alertado.
    const userIds = new Set<string>()
    if (params.channel_owner_id) userIds.add(params.channel_owner_id)

    const { data: admins } = await supa
        .from('tenant_members')
        .select('user_id')
        .eq('tenant_id', params.tenant_id)
        .in('role', ['clinic_admin', 'tenant_admin'])
        .eq('status', 'active')

    for (const a of admins || []) userIds.add(a.user_id)

    if (userIds.size === 0) return

    const kindLabel: Record<string, string> = {
        disconnected: 'desconectado',
        qr_pending: 'aguardando QR Code',
        error: 'com erro',
        connecting: 'tentando conectar',
        unknown: 'em estado desconhecido'
    }

    const title = `${params.channel_display} ${kindLabel[params.kind] || 'offline'}`
    const detail = `A conexão está fora há cerca de ${params.minutes_unhealthy} min. Envios automáticos podem estar falhando.`

    const rows = Array.from(userIds).map((uid) => ({
        owner_id: uid,
        tenant_id: params.tenant_id,
        type: 'system_alert',
        ref_id: params.incident_id,
        ref_table: 'whatsapp_connection_incidents',
        payload: {
            title,
            detail,
            severity: 'warn',
            deeplink: '/configuracoes/whatsapp-pessoal'
        }
    }))

    await supa.from('notifications').insert(rows)
    await supa.rpc('whatsapp_heartbeat_mark_notified', { p_incident_id: params.incident_id })
}

Deno.serve(async (req) => {
    if (req.method === 'OPTIONS') return new Response('ok', { headers: corsHeaders })

    const supa = createClient(
        Deno.env.get('SUPABASE_URL') ?? '',
        Deno.env.get('SUPABASE_SERVICE_ROLE_KEY') ?? '',
        { auth: { autoRefreshToken: false, persistSession: false } }
    )

    try {
        // Canal específico (on-demand via UI do tenant) ou varredura completa
        const url = new URL(req.url)
        const singleChannelId = url.searchParams.get('channel_id')

        let query = supa
            .from('notification_channels')
            .select('id, tenant_id, owner_id, provider, credentials, connection_status, last_health_check, metadata')
            .eq('provider', 'evolution_api')
            .eq('channel', 'whatsapp')
            .eq('is_active', true)
            .is('deleted_at', null)

        if (singleChannelId) query = query.eq('id', singleChannelId)

        const { data: channels, error: fetchErr } = await query

        if (fetchErr) return json({ error: fetchErr.message }, 500)
        if (!channels || channels.length === 0) {
            return json({ checked: 0, results: [] })
        }

        const now = new Date()
        const results = await Promise.all(
            channels.map((ch) => checkOneChannel(supa, ch as ChannelRow, now).catch((e) => ({
                tenant_id: (ch as ChannelRow).tenant_id,
                channel_id: (ch as ChannelRow).id,
                previous_status: (ch as ChannelRow).connection_status,
                new_status: 'error',
                action: 'fetch_error' as const,
                error: (e as Error).message
            })))
        )

        const summary = {
            checked: results.length,
            opened: results.filter((r) => r.action === 'opened').length,
            resolved: results.filter((r) => r.action === 'resolved').length,
            auto_reconnected: results.filter((r) => r.action === 'auto_reconnected').length,
            still_unhealthy: results.filter((r) => r.action === 'still_unhealthy').length,
            ok: results.filter((r) => r.action === 'ok').length,
            errors: results.filter((r) => r.action === 'fetch_error' || r.action === 'config_missing').length
        }

        return json({ ...summary, results })
    } catch (e) {
        return json({ error: (e as Error).message || 'unexpected_error' }, 500)
    }
})