Chat completions

POST

/api/v1/openai/chat/completions

One turn with the agent model names, in OpenAI's request and answer shapes, so an OpenAI SDK works with its base URL set to /api/v1/openai. The request carries the whole conversation and its last message is the user's. stream: true answers as chat.completion.chunk events ending with data: [DONE]. The agent's own tools run on SynapseOrch's side, so a request cannot bring tools. Errors are in OpenAI's shape. See OpenAI-compatible API.

Authorization

apiKey

headerAuthorizationBearer <token>

An API key, sk-syn- and 32 characters. On SynapseOrch Cloud an owner or admin makes it under Settings → Developer; on a self-hosted install, under Settings → API Keys.

Request Body

application/json

body

model?string

messages?array<>

Response Body

A chat completion; with stream true, server-sent chat.completion.chunk events ending with data: [DONE].

response

id?string

object?string

created?integer

model?string

choices?array<>

usage?

curl -X POST "https://api.synapseorch.com/api/v1/openai/chat/completions" \  -H "Authorization: Bearer sk-syn-..." \  -H "Content-Type: application/json" \  -d '{    "model": "Support Bot",    "messages": [      {        "role": "user",        "content": "Where is my order 4417?"      }    ]  }'

{  "id": "chatcmpl-5c1e9a7b3d2f4e6a8b0c1d2e",  "object": "chat.completion",  "created": 1726412345,  "model": "Support Bot",  "choices": [    {      "index": 0,      "message": {        "role": "assistant",        "content": "Order 4417 shipped on Monday; it should arrive by Thursday."      },      "finish_reason": "stop"    }  ],  "usage": {    "prompt_tokens": 812,    "completion_tokens": 21,    "total_tokens": 833  }}