Skip to main content

​
What Is a Provider

Each model on the platform may be backed by multiple providers. Providers are service nodes organized by source and cost — different providers have different pricing multipliers and model availability. When you send a request, the system automatically selects an available provider based on your token configuration. If a provider fails or times out, the system automatically retries another one — fully transparent to your code.

​
Available Providers

IDNameDescription
officialOfficialA provider group that aggregates direct connections to all major official platforms (OpenAI, Anthropic, Google, etc.); highest stability
official_cheapOfficial (Discounted)A cost-optimized version of official channels; high stability at a lower price
mixMixedMulti-source mixed channels; lowest price, widest model coverage

​
About official

official is a provider group that bundles nodes from OpenAI, Anthropic, Google, DeepSeek, and other major official platforms. When you select official, the system performs intelligent routing and failover across all official nodes — no need to specify each vendor individually.

​
How to Choose

Your needRecommended config
Maximum stability, price not a concernofficial
Good stability with better valueofficial_cheap
Lowest cost, minor instability acceptablemix
Try discounted official first, fall back to mixedofficial_cheap,mix

​
Configuring Providers in a Token

When creating or editing a token, you can set the default routing behavior for that token.

​
Provider List

Specifies which providers this token is allowed to use.
  • Leave empty (recommended): Automatically routes across all available providers for best availability
  • Single provider: e.g. official — routes only through official nodes
  • Multiple providers: Comma-separated, e.g. official_cheap,mix — tried in order or by sort strategy

​
Provider-Only Mode

  • Off (default): Specified providers are prioritized; falls back to others if all fail
  • On: Strictly limited to the specified providers — if all fail, returns an error with no fallback

​
Sort Strategy

When multiple providers are specified, controls the order they are tried:
  • Manual (default): Tried in the order you listed them
  • Latency: Prioritizes the fastest-responding provider
  • Price: Prioritizes the lowest-cost provider
  • Throughput: Prioritizes the highest-throughput provider

​
Overriding Providers Per Request

You can also temporarily override token settings for a single request via headers — no token changes needed.
HeaderEffect
X-Provider-OrderSpecifies the providers and order for this request
X-Provider-Only"true" enables strict mode — no fallback
X-Provider-SortOverrides the sort strategy (latency / price / throughput)
X-Provider-IgnoreProviders to skip for this request, stacked on top of token config
Providers specified via headers are still subject to your account permissions.

​
Examples

Try discounted official first, fall back to mixed on failure 
curl https://api.ephone.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-your-token" \
  -H "X-Provider-Order: official_cheap,mix" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-4o", "messages": [{"role": "user", "content": "Hello"}]}'
Strict official-only, no fallback allowed 
curl https://api.ephone.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-your-token" \
  -H "X-Provider-Order: official" \
  -H "X-Provider-Only: true" \
  -H "Content-Type: application/json" \
  -d '{"model": "claude-opus-4-5-20251101", "messages": [{"role": "user", "content": "Hello"}]}'
Set default headers in your SDK client 
import openai

client = openai.OpenAI(
    api_key="sk-your-token",
    base_url="https://api.ephone.ai/v1",
    default_headers={
        "X-Provider-Order": "official_cheap,mix",
    }
)