Pular para o conteúdo principal

Compatibility with OpenAI

Maritaca API is compatible with OpenAI client libraries, making it easy to experiment with our open-source models in existing applications.

This means that Sabiá models can be used in any program that uses OpenAI libraries.

Setting Up OpenAI to Use Maritaca API​

Library Installation​

First, install the openai library by typing this command in the terminal:

pip install openai

Setting Up the Client​

Setting up the OpenAI client is the first step to using the API. Make sure to provide your API key and the custom base URL. That is, to use Maritaca API, simply point the endpoint to https://chat.maritaca.ai/api, fill in the API key with a key obtained from the platform (as described in quick start), and use one of the Sabiá models.

import os
import openai

client = openai.OpenAI(
    api_key=os.environ.get("MARITACA_API_KEY"),
    base_url="https://chat.maritaca.ai/api",
)

Making a Request​

You can make a request to the sabia-4 model using the Responses API.

response = client.responses.create(
  model="sabia-4",
  instructions="You are a travel agent. Be descriptive and kind.",
  input="Tell me about the Christ the Redeemer.",
  max_output_tokens=8000
)

print(response.output[0].content[0].text)

Making a Request with Message List​

You can also pass a list of messages as input for multi-turn conversations.

response = client.responses.create(
  model="sabia-4",
  input=[
    {"role": "user", "content": "My name is Alice."},
    {"role": "assistant", "content": "Hello Alice! Nice to meet you."},
    {"role": "user", "content": "What is my name?"},
  ],
  max_output_tokens=8000
)

print(response.output[0].content[0].text)

Streaming Request​

To receive responses in real time (streaming), you can use the parameter stream=True.

import os
import openai

stream = client.responses.create(
  model="sabia-4",
  instructions="You are a travel agent. Be descriptive and kind.",
  input="Tell me about the Christ the Redeemer.",
  stream=True,
  max_output_tokens=8000
)
for event in stream:
  if event.type == "response.output_text.delta":
    print(event.delta, end="", flush=True)

Also available: Chat Completions API​

response = client.chat.completions.create(
  model="sabia-4",
  messages=[
    {"role": "system", "content": "You are a travel agent. Be descriptive and kind."},
    {"role": "user", "content": "Tell me about the Christ the Redeemer."},
  ],
  max_tokens=8000
)

print(response.choices[0].message.content)