Integrasi AI ke SCRM

1. Ringkasan Arsitektur

┌─────────────────────────────┐         ┌─────────────────────────────┐
│   SCRM 7 Star (Chatroom)    │         │   Materi Trading (AI Brain) │
│                             │  HTTPS  │                             │
│  Nasabah ◄──► CS asli       │ ──────► │  RAG + Prompt + Materi      │
│                             │  POST   │  /api/v1/ai/suggest         │
│  ┌───────────────────────┐  │         │                             │
│  │ Kotak Saran AI ◄───────┼──┼─────────┤  Return: suggested answer   │
│  │ (CS klik → edit → kirim)│  │         │  + meta stage/attachment    │
│  └───────────────────────┘  │         └─────────────────────────────┘
└─────────────────────────────┘

2. Prasyarat di Materi Trading

Pastikan hal-hal ini sudah siap sebelum SCRM mulai memanggil API:

Tips: uji dulu di /tanya-ai atau AI Playground admin sebelum hubungkan ke SCRM.

3. Prasyarat di SCRM 7 Star

Fitur minimum yang harus ada di SCRM agar integrasi berjalan:

4. Buat Token API

1. Login admin → buka Token API
2. Klik Create, isi nama mis. SCRM Production
3. Set rate limit sesuai kebutuhan (default cukup untuk tim CS)
4. Salin token yang muncul sekali (format: mtt_xxxxxxxx...) — simpan di env SCRM, jangan commit ke git
5. Opsional: isi Allowed IPs kalau mau batasi hanya dari server SCRM

5. Endpoint API

Base URL: https://www.materi.tools7star.shop
Auth header (wajib): Authorization: Bearer mtt_<token>
Content-Type: application/json

6. Request & Response

Request Body (POST /ai/suggest)

{
  "question": "apakah ini penipuan?",
  "mode": "suggestion",
  "history": [
    { "role": "assistant", "content": "Halo Kak! Saya Agustina dari Materi Tools 😊" },
    { "role": "user", "content": "halo, ini apa ya?" },
    { "role": "assistant", "content": "Program E*Trade Data resmi via PT Mega Menara Mas Berjangka..." }
  ]
}

Response Sukses (HTTP 200)

{
  "success": true,
  "data": {
    "answer": "Wajar Kak hati-hati... Program E*Trade Data BUKAN penipuan karena...",
    "sources": [ { "id": 12, "title": "Legalitas", "url": "...", "similarity": 0.82 } ],
    "images": [],
    "meta": {
      "provider": "groq",
      "model": "llama-3.3-70b-versatile",
      "latency_ms": 1240,
      "stage": "objection",
      "needs_attachment": null,
      "needs_link": null,
      "detected_cs_name": "Agustina",
      "mode": "suggestion"
    }
  }
}

7. Alur Integrasi di SCRM

1. 1

   Nasabah kirim pesan

   Pesan masuk ke chatroom SCRM seperti biasa.

2. 2

   SCRM kumpulkan history

   Ambil N pesan terakhir (user + assistant) dari database chat session.

3. 3

   SCRM panggil API

   POST ke /api/v1/ai/suggest dengan question = pesan terbaru nasabah, mode = suggestion, history = riwayat.

4. 4

   Tampilkan di Kotak Saran

   Render data.answer di panel saran — JANGAN auto-kirim ke nasabah.

5. 5

   CS review & edit

   CS baca saran, klik "Pakai", edit nama/link/gambar sesuai kebutuhan.

6. 6

   CS kirim manual

   CS tekan send — bubble chat yang nasabah lihat = versi final dari CS, bukan AI mentah.

8. Signal Khusus untuk CS

Dalam mode suggestion, AI bisa menyertakan signal text yang perlu ditangani CS di UI SCRM:

9. Contoh Kode

cURL — Health Check

curl -X GET "https://www.materi.tools7star.shop/api/v1/ai/health" \
  -H "Authorization: Bearer mtt_TOKEN_ANDA_DISINI" \
  -H "Accept: application/json"

cURL — Minta Saran

curl -X POST "https://www.materi.tools7star.shop/api/v1/ai/suggest" \
  -H "Authorization: Bearer mtt_TOKEN_ANDA_DISINI" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "question": "apakah ini penipuan?",
    "mode": "suggestion",
    "history": [
      {"role": "assistant", "content": "Halo Kak! Saya Agustina dari Materi Tools."},
      {"role": "user", "content": "halo, ini apa ya?"}
    ]
  }'

PHP (backend SCRM)

$response = Http::withToken(env('MATERI_AI_TOKEN'))
    ->timeout(30)
    ->post('https://www.materi.tools7star.shop/api/v1/ai/suggest', [
        'question' => $customerMessage,
        'mode'     => 'suggestion',
        'history'  => $chatHistory, // [['role'=>'user','content'=>'...'], ...]
    ]);

if ($response->successful() && $response->json('success')) {
    $suggestion = $response->json('data.answer');
    $stage      = $response->json('data.meta.stage');
    // Tampilkan $suggestion di kotak saran SCRM
}

JavaScript (via backend proxy — jangan expose token di frontend)

// Panggil endpoint internal SCRM Anda, BUKAN langsung ke Materi Trading
const res = await fetch('/internal/ai-suggest', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ sessionId, question }),
});
const { suggestion, meta } = await res.json();
// Render suggestion di kotak saran chatroom

10. Error & Rate Limit

Response headers: X-RateLimit-Limit-Minute, X-RateLimit-Remaining-Minute, X-RateLimit-Limit-Day, X-RateLimit-Remaining-Day

11. Checklist Go-Live