API Documentation

Build with Komilion

Integrate intelligent model orchestration, tool execution, and voice capabilities into your apps.

API Base URL

Important: Use the correct endpoint for all API requests

https://www.komilion.com/api

This base URL is OpenAI SDK compatible. Your client will call/chat/completions automatically.

Recommended endpoints:

POST /chat/completions - OpenAI-compatible Chat Completions (recommended)
POST /chat - Legacy chat endpoint (Komilion flat JSON)
POST /neo/tasks - Agent Harness ("Guardian Mode") tasks
POST /voice/realtime - Realtime Voice (WebSocket config)
POST /voice/transcribe-meeting - Meeting Transcription (speaker diarization)
POST /voice/upload - Pre-signed upload for large audio files
POST /deepgram/stt - Speech-to-text (short-form)
POST /deepgram/tts - Text-to-speech

Policy (Cost/Quality)

Choose your driving style via policy or by using neo-mode/<policy>

model: "neo-mode/balanced"

Quick Start

Get started with Komilion API in under 2 minutes

Sign Up & Get API Key

Create your account at komilion.com/auth/signup, then go to Dashboard → API Keys to generate your ck_your-api-key-here key

Make Your First Request

Send a chat completion request and let our orchestrator select the best model

Start Saving

Monitor your savings and let Komilion optimize your AI costs automatically

New Capabilities

Tool Execution

Agents can run Python code, search the web, and generate images.

Realtime Voice ⭐

Bidirectional speech-to-speech with GPT. Sub-300ms latency.

Meeting Transcription ⭐

Transcribe long meetings with speaker diarization. Export as JSON, SRT, or VTT.

Speech-to-Text

Transcribe audio files with Deepgram Nova-2.

Text-to-Speech

Convert text to speech with Aura models.

API Reference

Interactive code examples in your favorite language

Authentication

Use your API key from Dashboard → API Keys as a Bearer token

# Using your API key (recommended)
curl -X POST https://www.komilion.com/api/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -d '{
    "model": "neo-mode",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

How to get your API key:

Sign up at komilion.com/auth/signup
Go to Dashboard → API Keys
Click "Generate New Key"
Copy your ck_your-api-key-here key

Neo Mode (Autopilot)

POST /api/chat/completions • One API call that routes to direct chat, tool-using workflows, or long-running harness jobs

Request

# Neo Mode (Autopilot) — Komilion decides workflow + best execution model(s)
curl -X POST https://www.komilion.com/api/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -d '{
    "model": "neo-mode/balanced",
    "messages": [
      {"role": "user", "content": "Search + summarize the latest advancements in fusion energy. Include sources."}
    ],
    "taskHint": "reasoning"
  }'

Parameters

model

Use neo-mode (or neo-mode/premium) to let Komilion route intelligently

taskHint

Optional: "writing" | "code-generation" | "reasoning" | "vision"

Direct Model (Bypass Neo)

POST /api/chat/completions • Use any specific model by its OpenRouter ID

Request

# Use a specific model (bypass Neo routing)
curl -X POST https://www.komilion.com/api/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -d '{
    "model": "anthropic/claude-sonnet-4-5",
    "messages": [{"role": "user", "content": "Explain monads simply"}]
  }'

Use any model from the model catalog by its full ID (e.g., anthropic/claude-sonnet-4-5). Pricing is pass-through from the provider + 25% platform fee.

Speech-to-Text

POST /api/deepgram/stt • Transcribe audio files or URLs

Request

curl -X POST https://www.komilion.com/api/deepgram/stt \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -F "audio=@voice_memo.m4a"

Text-to-Speech

POST /api/deepgram/tts • Convert text to lifelike audio

Request

curl -X POST https://www.komilion.com/api/deepgram/tts \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -d '{
    "text": "Hello, I am your AI assistant.",
    "model": "aura-asteria-en"
  }' --output response.wav

Meeting Transcription (NEW)

POST /api/voice/transcribe-meeting • Long-form audio with speaker diarization

Key Features:

Speaker diarization — identifies who said what
Word-level timestamps and confidence scores
AI-powered meeting summary (optional)
Export: JSON, plain text, SRT, VTT subtitle formats
Large file support via pre-signed upload (up to 500MB)
Cost: ~$0.0054/min (Deepgram Nova-2 + 25% markup)

Parameters

audio

File upload (FormData) or audioKey / audioUrl (JSON)

language

Language code (e.g., "en", "sv"). Default: auto-detect

summarize

Boolean — Generate AI summary of the meeting

format

"json" (default) | "text" | "srt" | "vtt"

Implementation

# Transcribe a meeting with speaker diarization
curl -X POST https://www.komilion.com/api/voice/transcribe-meeting \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -F "audio=@meeting.mp3" \
  -F "language=en" \
  -F "summarize=true"

# For large files (>4MB), use pre-signed upload:
# Step 1: Get upload URL
curl -X POST https://www.komilion.com/api/voice/upload \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -d '{"filename": "meeting.mp3", "contentType": "audio/mpeg"}'

# Step 2: Upload file directly to R2
curl -X PUT "<uploadUrl from step 1>" \
  -H "Content-Type: audio/mpeg" \
  --data-binary @meeting.mp3

# Step 3: Transcribe using the key
curl -X POST https://www.komilion.com/api/voice/transcribe-meeting \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -d '{"audioKey": "<audioKey from step 1>", "language": "en", "summarize": true}'

Realtime Voice (NEW)

POST /api/voice/realtime • Bidirectional speech-to-speech conversations via WebSocket

🎤 Talk to AI in real-time!

Sub-300ms latency
Natural conversation flow
Automatic voice activity detection
Function calling support

Policies (just pick one!)

premium

Best quality (~$0.15/min)

balanced

Good quality, cheaper (~$0.04/min) DEFAULT

frugal

Basic, cheapest (~$0.01/min)

Voices

alloy •ash •ballad •coral •echo •sage •shimmer •verse

Implementation

# 1. Get WebSocket configuration (just pick a policy!)
curl -X POST https://www.komilion.com/api/voice/realtime \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ck_your-api-key-here" \
  -d '{
    "policy": "premium",
    "voice": "alloy"
  }'

# Policies:
#   "premium"  - Best quality (~$0.15/min)
#   "balanced" - Good quality, cheaper (~$0.04/min) [DEFAULT]
#   "frugal"   - Basic, cheapest (~$0.01/min)

# 2. Connect to WebSocket URL from response
# 3. Send: {"type": "session.update", "session": <sessionConfig>}
# 4. Stream audio as: {"type": "input_audio_buffer.append", "audio": "<base64>"}

Audio Format:

24kHz sample rate, mono, PCM16
Base64 encoded when sending/receiving

Pinned Mode — Your Models, Auto-Upgraded

Pin specific models per task type. Komilion auto-upgrades within the same provider family when newer versions launch.

How it works

Use the Comparison Playground to test models side-by-side
Click “Pin” on the winner
Future requests of that task type use your pinned model
When a newer version drops (e.g., Claude Sonnet 4.5 → 5.0), Komilion upgrades automatically. Same provider, better model.

Setting pins via API

POST /api/advisor/accept
{
  "taskType": "coding",
  "modelId": "anthropic/claude-sonnet-4-5"
}

Your pinned models are visible in the Dashboard under Settings. Requires Pro plan or higher.

Advisor Mode — Weekly Optimization Emails

Get proactive recommendations based on your actual usage patterns.

What you receive

Weekly email digest (Mondays, 09:00 UTC)
Cost analysis: where you're overpaying
Model recommendations: better options for your task types
New model alerts: when a relevant model launches
Health warnings: when models you use have issues

Each recommendation includes

“Test side-by-side” button → opens Comparison Playground pre-filled with recommended models and your typical prompt
One-click accept → pins the recommended model

Advisor emails activate automatically once you've made 10+ API calls. Requires Pro plan or higher.

Error Responses

Komilion returns standard HTTP status codes

Code	Meaning
400	Bad request (invalid parameters)
401	Invalid or missing API key
402	Insufficient wallet balance
429	Rate limit exceeded
500	Server error (automatic failover attempted)

Example error response

{
  "error": {
    "message": "Insufficient balance. Please top up at komilion.com/dashboard",
    "type": "billing_error",
    "code": 402
  }
}

Rate limit headers included in every response:

X-RateLimit-Limit: 120
X-RateLimit-Remaining: 118
X-RateLimit-Reset: 1707500000

Rate Limits

Per API key limits

Tier	Requests/min	Notes
Free tier	30/min	$5 trial credits
Standard	120/min	Pay-as-you-go
High volume	Custom	Contact us

Rate limits apply per API key. If you need higher limits, reach out at support@komilion.com.

Explore Documentation

Quick Start

Get started in 5 minutes

SDK Reference

Full SDK documentation

API Reference

All REST endpoints

Cline Integration

Setup guide for Cline

Examples

Code samples

Ready to Build?

Start integrating Komilion API into your applications today

Build with Komilion

API Base URL

Policy (Cost/Quality)

Frugal

Balanced Mode (Default)

Premium

Quick Start

Sign Up & Get API Key

Make Your First Request

Start Saving

New Capabilities

Tool Execution

Realtime Voice ⭐

Meeting Transcription ⭐

Speech-to-Text

Text-to-Speech

API Reference

Authentication

Neo Mode (Autopilot)

Request

Parameters

Direct Model (Bypass Neo)

Request

Speech-to-Text

Request

Text-to-Speech

Request

Meeting Transcription (NEW)

Parameters

Implementation

Realtime Voice (NEW)

Policies (just pick one!)

Voices

Implementation

Pinned Mode — Your Models, Auto-Upgraded

How it works

Setting pins via API

Advisor Mode — Weekly Optimization Emails

What you receive

Each recommendation includes

Error Responses

Example error response

Rate Limits

Explore Documentation

Quick Start

SDK Reference

API Reference

Cline Integration

Examples

Ready to Build?