Class Reference

class SwitchAI(provider, model_name, api_key=None)

The SwitchAI client class.

Parameters:

• provider (str) – The name of the provider to use.

• model_name (str) – The name of the model to use.

• api_key (str | None) – The API key to use, if not set it will be read from the environment variable. Defaults to None.

chat(messages, temperature=1.0, max_tokens=None, tools=None, response_format=None, stream=False)

Sends a chat request to the AI model and returns the response.

Parameters:

• messages (List[str | dict | ChatResponse]) – A list of messages to send to the model.

• temperature (float | None) – Sampling temperature to use, between 0 and 2. Higher values like 0.8 will make

• random (the output more)

• deterministic. (while lower values like 0.2 will make it more focused and)

• max_tokens (int | None) – The maximum number of tokens to generate. Defaults to None.

• tools (List | None) – A list of tools the model may call.

• response_format (Type[BaseModel] | None) – An object specifying the format that the model must output.

• stream (bool | None) – Whether to stream the response.

Returns:

The response from the model.

Return type:

ChatResponse

embed(inputs)

Embeds the input text using the AI model.

Parameters:

inputs (str | Image | List[str | Image]) – The input text to embed. Can be a single string or a list of strings.

Returns:

The response from the model.

Return type:

TextEmbeddingResponse

generate_image(prompt, n=1)

Generate an image based on the provided prompt.

Parameters:

• prompt (str) – A text description of the desired image.

• n (int) – The number of images to generate.

Returns:

The response from the model.

Return type:

ImageGenerationResponse

transcribe(audio_path, language=None)

Convert speech to text.

Parameters:

• audio_path (str) – The path to the audio file.

• language (str | None) – The language of the audio file.

Returns:

The response from the model.

Return type:

TranscriptionResponse

class ChatResponse(*, id=None, message=None, tool_calls=None, usage=None, finish_reason=None)

Represents a chat response from the model, based on the provided input.

Parameters:

• id (str | None) – A unique identifier of the response.

• message (ChatMessage | None) – The generated message.

• tool_calls (List[ChatToolCall] | None) – A list of tool calls.

• usage (ChatUsage | None) – Usage statistics.

• finish_reason (str | None) – The reason the generation finished. This will be completed if the generation was successful, max_tokens if the maximum token limit was reached,``content_filter`` if the content filter blocked the response,``tool_calls`` if the model called a tool, or unknown if the reason is unknown.

class ChatMessage(*, role=None, content=None)

The generated chat message.

Parameters:

• role (str | None) – The role of the author of this message.

• content (str | None) – The content of the message.

class ChatUsage(*, input_tokens=None, output_tokens=None, total_tokens=None)

Usage statistics for a chat response.

Parameters:

• input_tokens (int | None) – The number of input tokens used.

• output_tokens (int | None) – The number of output tokens generated.

• total_tokens (int | None) – The total number of tokens used.

class ChatToolCall(*, id=None, function, type='function')

A chat tool call.

Parameters:

• id (str | None) – A unique identifier of the tool call.

• function (Function) – The function called.

• type (str) – The function type. Always “function”.

class Function(*, name, arguments)

The function called by the model.

Parameters:

• name (str) – The name of the function.

• arguments (Dict[str, Any]) – The arguments of the function.

class EmbeddingResponse(*, id=None, object=None, model=None, usage=None, embeddings)

Represents an embedding response from the model, based on the provided input.

Parameters:

• id (str | None) – A unique identifier of the response.

• object (str | None) – The object type.

• model (str | None) – The model used to generate the response.

• usage (EmbeddingUsage | None) – Usage statistics.

• embeddings (List[Embedding]) – A list of embeddings.

class EmbeddingUsage(*, input_tokens=None, total_tokens=None)

Usage statistics for an embedding response.

Parameters:

• input_tokens (int | None) – The number of input tokens used.

• total_tokens (int | None) – The total number of tokens used.

class Embedding(*, index, data)

An embedding vector representing the input text.

Parameters:

• index (int) – The index of the embedding in the list of embeddings.

• data (List[float]) – The embedding vector, which is a list of floats.

class TranscriptionResponse(*, text)

A transcription of an input audio.

Parameters:

text (str) – The transcribed text.

class ImageGenerationResponse(*, images)

Represents an image generation response from the model, based on the provided input.

Parameters:

images (List[Any]) – A list of generated images.

SuperClients

class Browser(client)

A superclient that extends a chat SwitchAI client to support websites fetching and analysis.

Parameters:

client (SwitchAI) – A SwitchAI client initialized with a chat model.

chat(messages, temperature=1.0, max_tokens=None, tools=None, response_format=None, stream=False)

Sends a chat request to the AI model and returns the response.

Parameters:

• messages (List[str | dict | ChatResponse]) – A list of messages to send to the model.

• temperature (float | None) – Sampling temperature to use, between 0 and 2. Higher values like 0.8 will make

• random (the output more)

• deterministic. (while lower values like 0.2 will make it more focused and)

• max_tokens (int | None) – The maximum number of tokens to generate. Defaults to None.

• tools (List | None) – A list of tools the model may call.

• response_format (Type[BaseModel] | None) – An object specifying the format that the model must output.

• stream (bool | None) – Whether to stream the response.

Returns:

The response from the model.

Return type:

ChatResponse

class Classifier(client, classes, task_description=None, multi_label=False)

A superclient that extends a chat SwitchAI client to support classification tasks. It can be used to classify text or images.

Parameters:

• client (SwitchAI) – A SwitchAI client initialized with a chat model.

• classes (List[str]) – The classes to classify the data into.

• task_description (str) – A description of the classification task.

• multi_label (bool) – Whether the classifier should support multi-label classification or single-label classification.

classify(data)

Classifies the given data.

Parameters:

data (str | Image | List[str | Image]) – The data to classify.

Returns:

The classification result(s).

Return type:

str | List[str]

class Illustrator(client)

The Illustrator superclient generates illustrations based on text descriptions.

Parameters:

client (SwitchAI) – A chat SwitchAI client.

generate_illustration(description, output_path, image_reference=None, max_revision_steps=0, editor_mode=False)

Generates an illustration based on the given description and saves it to the specified output path.

Parameters:

• description (str) – The description of the illustration.

• output_path (str) – The path where the illustration will be saved. The file format should be SVG.

• image_reference (str | bytes | Image) – An image reference to be used to generate the illustration.

• max_revision_steps (int) – The maximum number of revision steps allowed to improve the illustration.

• 0 (If set to)

• Otherwise (no revisions will be made.)

• the (the model will continue refining)

• illustration (illustration until it reaches the maximum number of revision steps or until the)

• satisfactory. (is considered)

• editor_mode (bool) – If True, allows the user to interactively edit the illustration.

class ImageRetriever(client, images_folder_path, embeddings_cache_path=None, batch_size=32)

A superclient to retrieve images similar to a query image or text.

Parameters:

• client (SwitchAI) – A SwitchAI client that supports text and image embedding.

• images_folder_path (str) – The path to the folder containing the images.

• embeddings_cache_path (str | None) – The path to the embeddings cache file, else a file named ‘embeddings_cache.json’ will be created in the images folder.

• batch_size (int | None) – The batch size to use when embedding images.

retrieve_images(query, similarity_metric='cosine', threshold=0.5)

Retrieve images similar to the query image or text.

Parameters:

• query (str | Image) – The query image or text.

• similarity_metric (str) – The similarity metric to use. Must be ‘cosine’ or ‘euclidean’.

• threshold (float) – The similarity threshold.

Returns:

A sorted dictionary containing the image filenames as keys and the similarity scores as values.

Return type:

Dict[str, float]