API Documentation

curl https://api.genon.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-svp-YOUR_KEY" \
  -F "model=openai/whisper-large-v3" \
  -F "file=@audio.mp3" \
  -F "language=ko" \
  -F "response_format=verbose_json" \
  -F "diarize=true" \
  -F "num_speakers=3"   # known headcount → exactly N speakers (omit to auto-detect)

from openai import OpenAI
client = OpenAI(api_key="sk-svp-YOUR_KEY", base_url="https://api.genon.ai/v1")
with open("audio.mp3", "rb") as f:
    r = client.audio.transcriptions.create(
        model="openai/whisper-large-v3",
        file=f,
        language="ko",                 # optional — blank auto-detects
        response_format="verbose_json",  # json | verbose_json | srt | vtt
        extra_body={
            "diarize": True,           # ours: speaker labels per segment
            "num_speakers": 3,         # known headcount → pin to exactly N (omit to auto-detect)
        },
    )
print(r.text)
# r.segments[i].speaker present when diarize=True

import fs from "node:fs";
import OpenAI from "openai";
const client = new OpenAI({ apiKey: "sk-svp-YOUR_KEY", baseURL: "https://api.genon.ai/v1" });
const r = await client.audio.transcriptions.create({
    model: "openai/whisper-large-v3",
    file: fs.createReadStream("audio.mp3"),
    language: "ko",
    response_format: "verbose_json",
    // @ts-expect-error non-standard field (ours): speaker labels per segment
    diarize: true,
    // @ts-expect-error non-standard field (ours): pin to exactly N speakers (omit to auto-detect)
    num_speakers: 3,
});
console.log(r.text);
// r.segments[i].speaker present when diarize=true

// response_format=verbose_json, diarize=true
{
  "task": "transcribe",
  "language": "ko",
  "duration": 8160.885,                  // 오디오 길이(초)
  "text": "이게 온프레미스로 전달될 때 ...",   // 전체 평문 전사
  "segments": [
    { "id": 0, "start": 3.17, "end": 23.99,
      "text": "이게 온프레미스로 전달될 때 ...", "speaker": "SPEAKER_06" },
    { "id": 1, "start": 24.03, "end": 30.93,
      "text": "그래서 반자동, 자동화되는 형태라고 보시겠습니다.", "speaker": "SPEAKER_01" }
  ],
  "usage": { "type": "duration", "seconds": 8160.885,
             "billing_units": { "audio_seconds": 8160.885, "audio_minutes": 136.01 } },
  "request_id": "01KVDK1XWDGTCHYY56G0KQ6AJN"
}

// 화자 라벨(segments[].speaker, "SPEAKER_00"..)은 diarize=true 일 때만 채워집니다.
// response_format=json  → { "text": "..." } 만 반환(segments 없음).
// response_format=srt|vtt → 자막 텍스트(본문이 자막 문자열).
// timestamp_granularities[]=word → 최상위 "words": [{ "word", "start", "end" }] 추가.

Returns the calling key's label, monthly usage, budget cap, remaining credit, and rate limit. Mirrors OpenRouter's key-introspection shape.

curl https://api.genon.ai/v1/auth/key -H "Authorization: Bearer sk-svp-YOUR_KEY"

# 200 OK
{
  "data": {
    "label": "prod-bot",
    "usage": 12.34,             # USD spent this month
    "limit": 50.0,              # USD cap (null if inherited)
    "limit_remaining": 37.66,   # org credit balance, USD
    "is_free_tier": false,
    "is_provisioning_key": false,
    "rate_limit": { "requests": 60, "interval": "1m" }
  }
}

Org-level credit balance and total usage. Use for dashboards that don't need per-key detail.

curl https://api.genon.ai/v1/credits -H "Authorization: Bearer sk-svp-YOUR_KEY"

# 200 OK
{ "data": { "total_credits": 999.50, "total_usage": 12.34 } }

After a streamed completion, fetch the final token counts, cost, and latency by the request_id returned in the x-request-id response header. Requires your API key, and only returns generations belonging to your organisation. The billing row is written asynchronously, so a just-finished request may 404 for a few seconds.

curl "https://api.genon.ai/v1/generation?id=01HFAKEULID01" \
  -H "Authorization: Bearer $GENON_API_KEY"

# 200 OK
{
  "data": {
    "id": "gen-01HFAKEULID01",
    "request_id": "01HFAKEULID01",
    "model": "google/gemma-4-26b-a4b-it",
    "created_at": "2026-07-30T03:28:47.519000+00:00",
    "provider_name": "genon-nhn",
    "tokens_prompt": 1000,
    "tokens_completion": 500,
    "native_tokens_cached": 0,
    "total_cost": 0.000295,            # USD dollars, not cents
    "usage": 0.000295,
    "latency": 2300,                   # ms
    "generation_time": 2300,
    "status": "ok"
  }
}

total_cost is USD dollars — here 1000 input + 500 output tokens at gemma-4's $0.12/$0.35 per 1M.

Non-streaming /v1/chat/completions responses now include usage.cost (USD float). Tools like Aider and Continue render this directly without an extra /generation round-trip.

{
  "id": "chatcmpl-...",
  "choices": [...],
  "usage": {
    "prompt_tokens": 1000,
    "completion_tokens": 500,
    "total_tokens": 1500,
    "prompt_tokens_details": { "cached_tokens": 0 },
    "cost": 0.000750                    # USD, computed from per-model pricing
  },
  "provider": "genon-nhn"
}