Chat Completions

OpenAI-compatible chat completions endpoint. Drop-in compatible with the official OpenAI SDKs (Python, JS, etc.) — point them at Ragen as the base_url and they'll work unchanged.

POST https://api.ragen.ai/v1/chat/completions

Authentication

Authorization: Bearer YOUR_API_KEY

Each API key is scoped to a specific project — its knowledge base is what the model will retrieve against. Create and manage keys under Settings → API Keys in the dashboard.

Request

Body

Field	Type	Required	Description
`messages`	`array`	Yes	1–100 messages in conversation order. See Message format.
`model`	`string`	No	Model id to use (e.g. `gpt-5.4`). Defaults to the organization's default model. List available models via `GET /v1/models`.
`temperature`	`number`	No	0–2. Overrides the organization default.
`max_tokens`	`integer`	No	Cap on generated tokens (1–32,000).
`stream`	`boolean`	No	When `true`, responds with a Server-Sent Events stream. Default `false`.
`stream_options`	`object`	No	Streaming options. Currently supports `include_usage: boolean`. See Streaming.

Message format

Each message is { role, content }:

Role	Purpose
`user`	The user's turn. The last user message is treated as the active question; earlier ones become conversation history.
`assistant`	Prior assistant responses. Included in conversation history.
`system`	Per-request system instructions — merged on top of the project's own instructions (project owner's intent first, then the caller's).

Response

Non-streaming

Returns a chat.completion object:

{
  "id": "chatcmpl-7a2b4c...",
  "object": "chat.completion",
  "created": 1744664400,
  "model": "gpt-5.4",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Our refund policy allows returns within 30 days..."
      },
      "finish_reason": "stop",
      "logprobs": null
    }
  ],
  "usage": {
    "prompt_tokens": 128,
    "completion_tokens": 42,
    "total_tokens": 170
  }
}

Streaming

When stream: true, returns text/event-stream with a sequence of chat.completion.chunk objects, terminated by data: [DONE]:

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","created":1744664400,"model":"gpt-5.4","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null,"logprobs":null}]}

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","created":1744664400,"model":"gpt-5.4","choices":[{"index":0,"delta":{"content":"Our "},"finish_reason":null,"logprobs":null}]}

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","created":1744664400,"model":"gpt-5.4","choices":[{"index":0,"delta":{"content":"refund policy..."},"finish_reason":null,"logprobs":null}]}

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","created":1744664400,"model":"gpt-5.4","choices":[{"index":0,"delta":{},"finish_reason":"stop","logprobs":null}]}

data: [DONE]

The first chunk carries delta.role: "assistant" (OpenAI convention).
Content chunks carry delta.content.
The final chunk before [DONE] has empty delta and finish_reason: "stop".

Including usage in streams

By OpenAI convention, usage is not emitted on streaming responses unless the caller opts in. Pass stream_options: { include_usage: true } to receive a trailing usage chunk:

{
  "id": "chatcmpl-...",
  "object": "chat.completion.chunk",
  "created": 1744664400,
  "model": "gpt-5.4",
  "choices": [],
  "usage": {
    "prompt_tokens": 128,
    "completion_tokens": 42,
    "total_tokens": 170
  }
}

The choices: [] signals that this chunk carries usage only, not content.

Examples

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(
    base_url="https://api.ragen.ai/v1",
    api_key="YOUR_API_KEY",
)

resp = client.chat.completions.create(
    model="gpt-5.4",
    messages=[
        {"role": "user", "content": "What is our refund policy?"},
    ],
)
print(resp.choices[0].message.content)

Python streaming

stream = client.chat.completions.create(
    model="gpt-5.4",
    messages=[{"role": "user", "content": "Summarize our onboarding process"}],
    stream=True,
    stream_options={"include_usage": True},
)

for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
    if chunk.usage:
        print(f"\n\nUsed {chunk.usage.total_tokens} tokens")

JavaScript / TypeScript (openai-node)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.ragen.ai/v1",
  apiKey: process.env.RAGEN_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "gpt-5.4",
  messages: [{ role: "user", content: "What is our refund policy?" }],
});
console.log(resp.choices[0].message.content);

cURL

curl -X POST https://api.ragen.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "messages": [
      {"role": "user", "content": "What is our refund policy?"}
    ]
  }'

Multi-turn conversation

The last user message is treated as the active question; earlier messages form the conversation context.

resp = client.chat.completions.create(
    model="gpt-5.4",
    messages=[
        {"role": "user", "content": "What is our refund policy?"},
        {"role": "assistant", "content": "We offer refunds within 30 days of purchase."},
        {"role": "user", "content": "What about digital products?"},
    ],
)

Per-request system prompt

resp = client.chat.completions.create(
    model="gpt-5.4",
    messages=[
        {"role": "system", "content": "Respond only in Polish."},
        {"role": "user", "content": "What is our refund policy?"},
    ],
)

System messages are merged on top of the project's own instructions — the project owner's instructions come first, then the caller's.

Error responses

Errors use the OpenAI error envelope so existing SDK error-handling keeps working:

{
  "error": {
    "message": "prompt too long",
    "type": "invalid_request_error",
    "code": "context_length_exceeded",
    "param": null
  }
}

Status	`type`	Meaning
400	`invalid_request_error`	Malformed body, validation failure, prompt too long
401	`authentication_error`	Missing or invalid API key
403	`permission_error`	Key deactivated or missing required scope
404	`not_found_error`	Project (assistant) not found
429	`rate_limit_error`	Rate limit exceeded
5xx	`api_error`	Upstream / internal error

Rate limits

Chat completions run a full RAG pipeline (vector search + rerank + LLM call), so they're on the expensive tier:

Scope	Limit
`POST /v1/chat/completions`	10 req/min per IP

Each streaming and non-streaming request counts identically against this budget. When the limit is hit the response is 429 rate_limit_error; back off with jitter.

Differences vs. `/v1/chat`

The ragen-native POST /v1/chat endpoint is a simpler interface (single content string + optional context) that predates the OpenAI-compatible one. Both are supported:

	`/v1/chat`	`/v1/chat/completions`
Format	Ragen-native JSON / SSE	OpenAI wire format
Multi-turn	No (single prompt)	Yes (messages array)
Model override	No	Yes (`model` field)
Temperature override	No	Yes
`max_tokens`	No	Yes
Page context injection	Yes (`context` field)	No — use `messages` instead
Usage in streams	No	Opt-in via `stream_options`

Use /v1/chat/completions for any new integration that benefits from the OpenAI SDK ecosystem. Keep /v1/chat for the embed widget and other existing ragen-native consumers.

Authentication​

Request​

Body​

Message format​

Response​

Non-streaming​

Streaming​

Including usage in streams​

Examples​

Python (OpenAI SDK)​

Python streaming​

JavaScript / TypeScript (openai-node)​

cURL​

Multi-turn conversation​

Per-request system prompt​

Error responses​

Rate limits​

Differences vs. /v1/chat​