Create Chat Completion

POST https://ed.link/api/v2/openai/chat/completions

Sends a Chat Completions request through the application's configured providers. The response is the provider's Chat Completions object, plus session_id. It is not wrapped in $data.

You must provide your application secret key. This request is not made in the scope of one integration or person. X-Scope is the audit log scope: an arbitrary string, up to 128 bytes, that groups related events for retrieval. It is not sent upstream.

model must be an exact id configured on the application. Any other id is rejected. Omit model to use the application's priority list. A configured model is tried first, then the rest of the list. Each model is tried with every active key for that provider, oldest key first. A failed hop is retried only when nothing has been written to the client yet.

stream: true returns server-sent events. Any other value returns one JSON completion. The OpenAI SDK posts to {baseURL}/chat/completions. Set baseURL to https://ed.link/api/v2/openai.

session_id and context are removed before the provider call. A body scope property is forwarded and is not the X-Scope header. Every other field is forwarded, including a property this API does not document. { "hello": "world" } is sent upstream as hello. A provider can still reject a field it does not support.

Headers

A blocked browser Origin returns 403.

Request Body

Edlink rejects a body that is not a JSON object, an empty messages array, a missing or oversized X-Scope, an oversized session id, a context value that is not null or { "head", "tail" }, and a system message whose content is not a string when several system messages are merged. The messages kept after trimming must be at most 7,999,522 bytes.

Sample Request

const response = await axios.post(
    'https://ed.link/api/v2/openai/chat/completions',
    {
        model: 'openai/gpt-4o',
        messages: [
            { role: 'system', content: 'You are a helpful assistant.' },
            { role: 'user', content: 'What is the capital of France?' },
            { role: 'assistant', content: 'The capital of France is Paris.' },
            { role: 'user', content: 'What river runs through it?' }
        ]
    },
    {
        headers: {
            authorization: `Bearer ${application_secret_key}`,
            'x-scope': 'acme-tenant-7f3a',
            'x-session-id': 'conv_123'
        }
    }
);

The same call with the OpenAI SDK. baseURL is https://ed.link/api/v2/openai. The SDK adds /chat/completions.

import OpenAI from 'openai';

const client = new OpenAI({
    apiKey: application_secret_key,
    baseURL: 'https://ed.link/api/v2/openai',
    defaultHeaders: {
        'X-Scope': 'acme-tenant-7f3a',
        'X-Session-Id': 'conv_123'
    }
});

const completion = await client.chat.completions.create({
    model: 'openai/gpt-4o',
    messages: [
        { role: 'system', content: 'You are a helpful assistant.' },
        { role: 'user', content: 'What is the capital of France?' }
    ]
});

Set stream to true for server-sent events. Each chunk includes session_id. The stream ends with data: [DONE].

const stream = await client.chat.completions.create({
    model: 'openai/gpt-4o',
    stream: true,
    messages: [
        { role: 'system', content: 'You are a helpful assistant.' },
        { role: 'user', content: 'What is the capital of France?' }
    ]
});

for await (const chunk of stream) {
    const content = chunk.choices[0]?.delta?.content;
    if (content) {
        process.stdout.write(content);
    }
}

Sample Response

session_id is also returned in the X-Session-Id response header.

{
    "id": "chatcmpl-123",
    "object": "chat.completion",
    "created": 1677652288,
    "model": "openai/gpt-4o",
    "choices": [
        {
            "index": 0,
            "message": {
                "role": "assistant",
                "content": "The Seine runs through Paris."
            },
            "finish_reason": "stop"
        }
    ],
    "usage": {
        "prompt_tokens": 25,
        "completion_tokens": 10,
        "total_tokens": 35
    },
    "session_id": "conv_123"
}