API reference
Generated from the API's OpenAPI spec (the public /v1 part only). Base URL: https://api.beatriceai.it. For usage details see the pages of the API section.
GET /v1/models
List available models
200 response
| Name | Type | Notes |
|---|---|---|
object | string | |
data | object[] |
GET /v1/models/{model}
Retrieve a model
Parameters
| Name | Type | Notes |
|---|---|---|
model (path) | string | required |
200 response
| Name | Type | Notes |
|---|---|---|
id | string | |
object | string | |
created | integer | |
owned_by | string |
POST /v1/chat/completions
Create a chat completion (OpenAI compatible, streaming supported)
OpenAI chat completions format. The answer may carry reasoning_content next to content; usage counts reasoning in completion_tokens. A red-line refusal is a normal answer with finish_reason: "content_filter".
Request body
Also forwarded: stop, presence_penalty, frequency_penalty, seed, response_format, tools, tool_choice, parallel_tool_calls, logprobs, top_logprobs, reasoning_effort. Other fields are ignored.
| Name | Type | Notes |
|---|---|---|
model | string | required. Model id, e.g. beatrice-flash (GET /v1/models) |
messages | object[] | required. The conversation so far |
stream | boolean | Send the answer as server-sent events |
stream_options | object | |
max_tokens | integer | Output limit, reasoning included; default 8192 · ≥ 1 |
max_completion_tokens | integer | Same as max_tokens (newer OpenAI name) · ≥ 1 |
n | integer | Only 1 is supported · ≥ 1 |
temperature | number | ≥ 0, ≤ 2 |
top_p | number | ≥ 0, ≤ 1 |