OpenAI Chat Completions Compatible Mode

/v1/chat/completions is the de facto standard interface of the LLM industry — virtually every framework, client, and SDK supports it out of the box. Through APIYI, this single endpoint reaches OpenAI, Claude, Gemini, DeepSeek, and 400+ models in total; switching models is just swapping a string.

Which endpoint to pick: using existing frameworks/clients, or want one codebase across multiple vendors → compatible mode (this page); need built-in tools (web search, code interpreter), state management, or Pro-series models → Native Calls (/v1/responses). OpenAI’s official stance on Chat Completions: supported long-term, but Responses is recommended for new projects.

Quick Start

curl https://api.apiyi.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "gpt-5.4",
    "messages": [
      {"role": "user", "content": "Introduce yourself in one sentence"}
    ]
  }'

One Interface, Every Provider

This is the biggest payoff of compatible mode: switching models means changing a string — not a line of code.

def ask(message: str, model: str) -> str:
    response = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": message}]
    )
    return response.choices[0].message.content

print(ask("Explain quantum entanglement", "gpt-5.4"))               # OpenAI
print(ask("Explain quantum entanglement", "claude-sonnet-4-6"))      # Anthropic
print(ask("Explain quantum entanglement", "gemini-3-pro-preview"))   # Google
print(ask("Explain quantum entanglement", "deepseek-chat"))          # DeepSeek

Full model names and prices: Models & Pricing. Note: calling Claude through the compatible format forfeits Claude’s Prompt Cache discount — for heavy Claude usage, use Claude Native Calls.

SDK Setup per Language

Every official SDK supports a custom base_url — configure once and go.

Python

pip install openai

from openai import OpenAI, AsyncOpenAI

# Synchronous client
client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.apiyi.com/v1"
)

# Async client
async_client = AsyncOpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.apiyi.com/v1"
)

Or use environment variables for zero in-code config:

export OPENAI_API_KEY="YOUR_API_KEY"
export OPENAI_BASE_URL="https://api.apiyi.com/v1"

from openai import OpenAI
client = OpenAI()  # reads the environment automatically

Node.js / TypeScript

npm install openai

import OpenAI from 'openai';

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  baseURL: 'https://api.apiyi.com/v1'
});

const response = await openai.chat.completions.create({
  model: 'gpt-5.4-mini',
  messages: [{ role: 'user', content: 'Hello!' }],
  temperature: 0.7
});

.NET

dotnet add package OpenAI

using OpenAI;
using OpenAI.Chat;

var client = new OpenAIClient(
    new System.ClientModel.ApiKeyCredential("YOUR_API_KEY"),
    new OpenAIClientOptions { Endpoint = new Uri("https://api.apiyi.com/v1") }
);

var chatClient = client.GetChatClient("gpt-5.4");
var response = await chatClient.CompleteChatAsync("Hello!");
Console.WriteLine(response.Value.Content[0].Text);

Go

Use the official OpenAI Go SDK (github.com/openai/openai-go):

go get github.com/openai/openai-go

package main

import (
    "context"
    "fmt"

    "github.com/openai/openai-go"
    "github.com/openai/openai-go/option"
)

func main() {
    client := openai.NewClient(
        option.WithAPIKey("YOUR_API_KEY"),
        option.WithBaseURL("https://api.apiyi.com/v1"),
    )

    completion, err := client.Chat.Completions.New(context.TODO(), openai.ChatCompletionNewParams{
        Model: "gpt-5.4",
        Messages: []openai.ChatCompletionMessageParamUnion{
            openai.UserMessage("Hello!"),
        },
    })
    if err != nil {
        panic(err)
    }
    fmt.Println(completion.Choices[0].Message.Content)
}

Java

Use the official OpenAI Java SDK (com.openai:openai-java):

<dependency>
    <groupId>com.openai</groupId>
    <artifactId>openai-java</artifactId>
    <version>LATEST</version>
</dependency>

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.chat.completions.ChatCompletion;
import com.openai.models.chat.completions.ChatCompletionCreateParams;

OpenAIClient client = OpenAIOkHttpClient.builder()
    .apiKey("YOUR_API_KEY")
    .baseUrl("https://api.apiyi.com/v1")
    .build();

ChatCompletionCreateParams params = ChatCompletionCreateParams.builder()
    .model("gpt-5.4")
    .addUserMessage("Hello!")
    .build();

ChatCompletion completion = client.chat().completions().create(params);
System.out.println(completion.choices().get(0).message().content().orElse(""));

Legacy projects on third-party libraries (Go’s sashabaranov/go-openai, Java’s theokanning packages) still work after changing the base_url, but we recommend migrating to the official SDKs above — third-party libraries lag on new parameters such as reasoning_effort.

Common Features

Streaming

stream = client.chat.completions.create(
    model="gpt-5.4",
    messages=[{"role": "user", "content": "Write a short poem about autumn"}],
    stream=True
)

for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

Reasoning control

On Chat Completions, use the top-level reasoning_effort parameter (different from the nested form on Responses):

response = client.chat.completions.create(
    model="gpt-5.4",
    messages=[{"role": "user", "content": "Prove that the square root of 2 is irrational"}],
    reasoning_effort="high"  # none / low / medium / high / xhigh
)

gpt-5 series reasoning models do not support temperature / top_p on this endpoint either — passing them raises an error.

Image input

response = client.chat.completions.create(
    model="gpt-5.4",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "What's in this image?"},
                {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}}
            ]
        }
    ]
)

Embeddings

response = client.embeddings.create(
    model="text-embedding-3-small",
    input="Text to embed"
)
embedding = response.data[0].embedding

Error Handling and Retries

The official SDKs retry automatically (2 attempts by default, on 429 / 5xx / connection errors) — prefer that over hand-rolled loops:

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.apiyi.com/v1",
    max_retries=3,   # built-in exponential backoff
    timeout=60.0
)

For finer control, catch by exception type:

from openai import (
    APIError,
    APIConnectionError,
    RateLimitError,
    InternalServerError,
)

try:
    response = client.chat.completions.create(
        model="gpt-5.4",
        messages=[{"role": "user", "content": "Hello"}]
    )
except RateLimitError:
    print("Rate limited — retry later")
except APIConnectionError:
    print("Connection error — check network/proxy")
except InternalServerError:
    print("Upstream error — worth retrying")
except APIError as e:
    print(f"API error: {e}")

Capability Boundaries of Compatible Mode

Capability	Compatible mode	Notes
Chat / streaming / multimodal input	✅	Fully supported
Function calling (FC)	✅	See Function Calling
Prompt cache discount	✅	Automatic for OpenAI models — see Cache Billing
Built-in tools (web search, code interpreter, …)	❌	Native Calls only
State management (previous_response_id)	❌	Native only; here you assemble history yourself
`verbosity` output control	❌	Native only
Pro-series models (gpt-5.4-pro, …)	❌	In practice, native calls only

Migrating from OpenAI Direct

Already on OpenAI’s official service? Migration is two steps with zero code changes:

Change base_url and key

# Before
client = OpenAI(api_key="sk-...")

# After
client = OpenAI(
    api_key="YOUR_APIYI_KEY",
    base_url="https://api.apiyi.com/v1"
)

Or change environment variables only (code untouched)

export OPENAI_API_KEY="YOUR_APIYI_KEY"
export OPENAI_BASE_URL="https://api.apiyi.com/v1"

Method calls, parameter formats, and response structures all stay identical.

This group: Native Calls · Cache Billing · Function Calling
Models & pricing: Models & Pricing
Get / manage tokens: https://api.apiyi.com/token
Official OpenAI SDK list: platform.openai.com/docs/libraries

​Quick Start

​One Interface, Every Provider

​SDK Setup per Language

​Python

​Node.js / TypeScript

​.NET

​Go

​Java

​Common Features

​Streaming

​Reasoning control

​Image input

​Embeddings

​Error Handling and Retries

​Capability Boundaries of Compatible Mode

​Migrating from OpenAI Direct

​Related Links

Quick Start

One Interface, Every Provider

SDK Setup per Language

Python

Node.js / TypeScript

.NET

Go

Java

Common Features

Streaming

Reasoning control

Image input

Embeddings

Error Handling and Retries

Capability Boundaries of Compatible Mode

Migrating from OpenAI Direct

Related Links