Sawalni Python SDK

This is the official Python SDK for the Sawalni API, providing easy access to language-related services such as embedding generation, language identification, and translation. Sawalni API is developed by Omneity Labs, and provides unique multilingual models and NLP capabilities, including pioneering Moroccan Darija support.

Installation

Install the package using pip:

pip install sawalni

Quick Start

To use the Sawalni SDK, you'll need an API key. You can set it as an environment variable or pass it directly to the client:

from sawalni import Sawalni

client = Sawalni(api_key='your_api_key_here') 
# or specify the key via SAWALNI_API_KEY in the environment

Features

The SDK supports both synchronous and asynchronous operations for the following services:

1. Chat: Generate chat completions using our Sawalni multilingual models, supporting Moroccan Darija, English, French, Arabic and many other languages.
2. Search: Perform internet searches in multiple languages with a single query.
3. Generate Embeddings: Create embeddings for text in multiple languages using Madmon.
4. Identify Language: Detect the language of a given text with Gherbal, supporting up to 33 languages.
5. Translate Text: Translate text between 13 supported languages with Tarjamli.
6. Transliterate Text: Convert Moroccan Arabic script to Moroccan Latin script with Daktilo.

The Sawalni SDK includes an OpenAI compatible client, which can be accessed via the chat and embeddings properties, or direct use via the OpenAI client as detailed below.

Chat

# Available models: sawalni-micro, sawalni-mini, sawalni-small
chat = client.chat.completions.create(messages=[{"role": "user", "content": "Hello, how are you?"}], model="sawalni-small")

# Stream
stream = client.chat.completions.create(messages=[{"role": "user", "content": "Hello, how are you?"}], model="sawalni-small", stream=True)
for chunk in stream:
    print(chunk.choices[0].delta.content)

Search

search = client.search("Hello, how are you?")

Generate Embeddings

embeddings = client.embed("Hello, world!")

Identify Language

language = client.identify("Bonjour le monde")

Translate Text

# You can specify a source language or let the model detect it automatically
translation = client.translate("Hello", source="auto", target="ary_Latn")

Transliterate Text

transliteration = client.transliterate("اهلا بيك", model="daktilo-mini", to="latn", temperature=0.1)

# {"text": "ahlane bik"}

Asynchronous Usage

For asynchronous operations, use the SawalniAsync client:

from sawalni import SawalniAsync

async_client = SawalniAsync(api_key='your_api_key_here')
embeddings = await async_client.embed("Hello, world!")

OpenAI compatible client

The SDK also includes an OpenAI compatible client, which can be accessed via the chat and embeddings properties:

chat = client.chat
embeddings = client.embeddings

You can also use the OpenAI client directly with the base URL set to https://api.sawalni.com/v1 and the API key set to your Sawalni API key.

import openai
client = openai.OpenAI(base_url="https://api.sawalni.com/v1", api_key="your_api_key_here")

Only the chat and embeddings properties are supported with this approach.

Documentation

For detailed information about available models, parameters, languages and and response formats, please refer to the complete API documentation at https://api.sawalni.com.

Support

If you encounter any issues or have questions, please contact api@sawalni.com.