Developer Documentation

Welcome to ArkAgentic Developer API docs. Integrate global SOTA models with one endpoint.

1. Quickstart

Use your existing OpenAI SDK. Only change base URL tohttps://gateway.arkagentic.com/v1and pass your ArkAgentic API key.

Get API Key βž”

cURL Β· bash

curl https://gateway.arkagentic.com/v1/chat/completions \
  -H "Authorization: Bearer ***" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "ark-gpt-4o",
    "messages": [{"role":"user","content":"Hello from ArkAgentic"}],
    "temperature": 0.2,
    "stream": false
  }'

2. Model Routing Directory

Pass standard ArkAgentic model IDs in model. Gateway can map aliases and upstream overrides internally while your client integration stays stable.

Model IDCategoryAlias / Route TargetNotes
ark-claude-3-5-sonnetAnthropicclaude-3-5-sonnetGeneral production chat workload
ark-claude-opus-5Anthropicclaude-opus-5General production chat workload
ark-claude-sonnet-5Anthropicclaude-sonnet-5General production chat workload
ark-cohere-embed-v3OpenAICohere-embed-v3-multilingualVector embedding and semantic indexing
ark-cohere-rerank-v4-fastOpenAICohere-rerank-v4.0-fastRetrieval reranking and relevance scoring
ark-cohere-rerank-v4-proOpenAICohere-rerank-v4.0-proRetrieval reranking and relevance scoring
ark-deepseek-r1DeepSeekdeepseek-reasonerDeep reasoning and long-chain analysis tasks
ark-deepseek-v4-flashDeepSeekDeepSeek-V4-FlashGeneral production chat workload
ark-deepseek-v4-proDeepSeekDeepSeek-V4-ProGeneral production chat workload
ark-gemini-1.5-proOpenAIgemini-1.5-proGeneral production chat workload

For detailed token pricing and compute estimator, visitPricing βž”.

View all 19 models in Model Center βž”
Page 1 / 2

3. Advanced Gateway Features

Failover & Auto-Retry

If an upstream node is degraded or unavailable, gateway automatically reroutes and retries to healthy candidates with no client-side route change.

Streaming SSE

Standard OpenAI-compatible streaming is supported. Setstream: trueand consume Server-Sent Events in your existing client.

Streaming Request

{
  "model": "ark-gpt-4o",
  "messages": [{"role":"user","content":"stream this response"}],
  "stream": true
}

HTTP Status Codes & Error Handling

CodeMeaningSuggested Action
400Invalid request payloadValidate JSON shape and model ID.
401UnauthorizedCheck Bearer token and key scope.
429Rate limitedApply exponential backoff and retry.
5xxTransient upstream/gateway failureRetry with jitter; gateway may auto-failover.