API

# Chat Completions

OpenAI Chat Completions. Point an OpenAI SDK at `https://inference.dagon.so/v1` and call it as usual.

**On this page**

- [Request](https://inference.dagon.so/docs/chat-completions#request)
- [Forwarded fields](https://inference.dagon.so/docs/chat-completions#forwarded-fields)
- [Response](https://inference.dagon.so/docs/chat-completions#response)

## Request

POST `/v1/chat/completions`

**curl**

```curl
curl https://inference.dagon.so/v1/chat/completions \
  -H "Authorization: Bearer $DAGON_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "abliterated-model-large-flash",
    "messages": [{"role": "user", "content": "Explain this stack trace."}],
    "max_tokens": 1024
  }'
```

**Python**

```python
# pip install openai
import os
from openai import OpenAI

client = OpenAI(base_url="https://inference.dagon.so/v1", api_key=os.environ["DAGON_KEY"])

reply = client.chat.completions.create(
    model="abliterated-model-large-flash",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
    max_tokens=1024,
)
print(reply.choices[0].message.content)
```

**TypeScript**

```ts
// npm install openai
import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://inference.dagon.so/v1", apiKey: process.env.DAGON_KEY });

const reply = await client.chat.completions.create({
  model: "abliterated-model-large-flash",
  messages: [{ role: "user", content: "Explain this stack trace." }],
  max_tokens: 1024,
});
console.log(reply.choices[0].message.content);
```

## Forwarded fields

Only documented fields whose cost the meter can see are forwarded: `model`, `messages`, `max_tokens`, `max_completion_tokens`, `n`, `stream`, `stream_options`, `temperature`, `top_p`, `stop`, `presence_penalty`, `frequency_penalty`, `logit_bias`, `logprobs`, `top_logprobs`, `seed`, `response_format`, `tools`, `tool_choice`, `parallel_tool_calls`, `functions`, `function_call`, `reasoning_effort`, `verbosity`, `prompt_cache_key`, `store`, `metadata`, `user`, `safety_identifier`. Any other non-null field is `400 unsupported_parameter` (null fields are dropped).

- **Output cap.** `max_completion_tokens` wins over `max_tokens`, and only the checked one is forwarded. Without either, 16,384 is written for you.
- **`n`**: 1 to 8, chat only. It multiplies the hold.
- **Tools.** Only client-executed `function` (or `custom`) tools. `tool_choice` may be `none`, `auto`, `required` or name one of the request's own functions.
- **Images.** `image_url` parts with a `data:image/…;base64,` URL or an http(s) URL, on models whose `input_modalities` include `image`.

Not available

Server-run tools (web search, code execution, file search, MCP, image generation), audio output, `prediction`, `service_tier` and content stored upstream (`file_id`) are refused: none can be priced per token. Web search is `400 web_search_not_supported`.

## Response

The provider's OpenAI-format body, unchanged, with these headers added:

| Header | Value |
| --- | --- |
| `x-gateway-request-id` | The gateway's id for this request; quote it to support. |
| `x-gateway-receipt-id` | The receipt id; fetch it at `GET /v1/receipts/:id`. |
| `x-gateway-receipt` | Non-streaming only: the signed receipt itself, base64url JSON. |

Reading headers with the SDKs:

**Python**

```python
raw = client.chat.completions.with_raw_response.create(
    model="abliterated-model-large-flash", messages=[{"role": "user", "content": "Hi"}], max_tokens=200,
)
print(raw.headers.get("x-gateway-receipt-id"))
reply = raw.parse()
```

**TypeScript**

```ts
const { data, response } = await client.chat.completions
  .create({ model: "abliterated-model-large-flash", messages: [{ role: "user", content: "Hi" }], max_tokens: 200 })
  .withResponse();
console.log(response.headers.get("x-gateway-receipt-id"));
```

Last updated 2026-11-07

[Previous **Passkey accounts**](https://inference.dagon.so/docs/passkeys) [Next **Responses**](https://inference.dagon.so/docs/responses)