Dokumentasi — Arus API

1. Autentikasi

Semua request pakai header Authorization: Bearer <api_key>.

Authorization: Bearer arus_...

2. Chat completion

POST /v1/chat/completions

{
  "model": "deepseek-v4-pro",
  "messages": [
    {"role": "system", "content": "Kamu asisten."},
    {"role": "user", "content": "Halo"}
  ],
  "temperature": 0.7,
  "max_tokens": 1024
}

3. Streaming

stream = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[...],
    stream=True,
)

4. Tool calling

tools = [{"type": "function",
          "function": {"name": "get_weather",
                       "parameters": {...}}}]

5. Python lengkap

from openai import OpenAI
client = OpenAI(base_url="https://api.router.trackster.dev/v1",
                api_key="arus_...")
r = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[{"role": "user", "content": "Halo"}],
)
print(r.choices[0].message.content)

6. Model tersedia

Model asli, bukan rebrand. Cek di GET /v1/models.

deepseek-v4-pro deepseek-v4.1-flash deepseek-v3.2 glm-5.3 glm-5.3-flash glm-5.2 glm-5v-turbo kimi-k2.7 kimi-k2.6 kimi-k3 minimax-m3 hy4-preview hy3

7. Thinking / Reasoning

Kirim "reasoning_effort": "high" di body request buat nampilin proses berpikir (reasoning_content di stream). Tanpa itu model langsung jawab tanpa reasoning.

{
  "model": "deepseek-v4-pro",
  "reasoning_effort": "high",
  "messages": [...]
}

8. Tabel model

Model	Context	Reasoning	Vision	Tool calling	Harga per 1M
deepseek-v4-pro	128k	✅ via reasoning_effort	—	✅	Rp 100
deepseek-v4.1-flash	128k	—	✅	✅	Rp 100
glm-5.3	128k	✅	—	✅	Rp 100
glm-5.3-flash	128k	—	✅	✅	Rp 100
glm-5.2	128k	—	—	✅	Rp 100
glm-5v-turbo	128k	—	✅	✅	Rp 100
kimi-k2.7	128k	✅	✅	✅	Rp 100
kimi-k2.6	128k	—	✅	✅	Rp 100
kimi-k3	128k	✅	✅	✅	Rp 100
minimax-m3	128k	—	✅	✅	Rp 100
hy4-preview	128k	✅	—	✅	Rp 100
hy3	128k	✅ (tanpa jawaban)	—	—	Rp 100

Harga flat Rp 100 per 1M token (input + output dihitung sama). Reasoning tidak ditagih sebagai output.

9. GET /v1/models

curl https://api.router.trackster.dev/v1/models \
  -H "Authorization: Bearer arus_..."

Balasan: daftar id model asli yang tersedia.

10. Rate limit & kode error

Rate limit per API key: 60 RPM · per IP: 180 RPM. Format error JSON:

{ "error": { "message": "...", "type": "quota|rate_limit" } }

401 — invalid api key.

402 — saldo token habis.

429 — rate limit.

502 — semua akun upstream gagal (coba lagi).

11. Tool calling lengkap

Langkah 1 — kirim definisi tool:

{"model":"deepseek-v4-pro",
 "messages":[{"role":"user","content":"Cuaca Jakarta?"}],
 "tools":[{"type":"function","function":{"name":"get_weather","parameters":{"type":"object","properties":{"city":{"type":"string"}}}}}]}

Langkah 2 — balasan berisi tool_calls. Kirim balik hasilnya:

"messages": [
  ...,
  {"role":"assistant","tool_calls":[{"id":"call_1","type":"function","function":{"name":"get_weather","arguments":"{\"city\":\"Jakarta\"}"}}]},
  {"role":"tool","tool_call_id":"call_1","content":"Hujan ringan, 27°C"}
]

Langkah 3 — kirim ulang, model kasih jawaban final.

12. Monitoring saldo & usage

curl https://api.router.trackster.dev/api/me \
  -H "Cookie: cbcn_session=..."

curl https://api.router.trackster.dev/api/me/usage/daily \
  -H "Cookie: cbcn_session=..."

Status layanan: /status