AxiomAi (AxiomAI v0.1.13)

View Source

AxiomAI - A unified Elixir client for multiple AI providers.

This library provides a consistent interface for interacting with various AI providers including Vertex AI, OpenAI, Anthropic, and local PyTorch models.

Example

iex> client = AxiomAi.new(:vertex_ai, %{project_id: "my-project", region: "us-central1"})
iex> AxiomAi.chat(client, "Hello, how are you?")
{:ok, %{response: "I'm doing well, thank you for asking!"}}

Summary

Types

client()
config()
message()
provider()
response()

Functions

batch_embed(client, texts, opts \\ %{})

Generates embeddings for multiple texts in a single API call.

chat(client, message)

Sends a chat message to the AI provider.

chat(client, system_prompt, history, prompt)

Sends a chat message with system prompt, history, and user prompt.

complete(client, prompt, options \\ %{})

Generates completions based on a prompt.

embed(client, text, opts \\ %{})

Generates an embedding vector for a single text.

new(provider, config \\ %{})

Creates a new client for the specified AI provider.

stream(client, message)

Streams a chat message to the AI provider.

stream(client, system_prompt, history, prompt)

Streams a chat message with system prompt, history, and user prompt.

Types

client()

@type client() :: %AxiomAi.Client{config: term(), provider: term()}

config()

@type config() :: map()

message()

@type message() :: String.t()

provider()

@type provider() :: :vertex_ai | :openai | :anthropic | :local

response()

@type response() :: {:ok, map()} | {:error, any()}

Functions

batch_embed(client, texts, opts \\ %{})

@spec batch_embed(client(), list(), map()) :: response()

Generates embeddings for multiple texts in a single API call.

Examples

iex> client = AxiomAi.new(:vertex_ai, %{project_id: "my-project"})
iex> {:ok, %{embeddings: vectors}} = AxiomAi.batch_embed(client, ["text 1", "text 2"])
iex> length(vectors)
2

chat(client, message)

@spec chat(client(), message()) :: response()

Sends a chat message to the AI provider.

Parameters

  • client: The client instance
  • message: The message to send

Examples

iex> client = AxiomAi.new(:vertex_ai, %{project_id: "my-project"})
iex> AxiomAi.chat(client, "What is the weather like?")
{:ok, %{response: "I don't have access to real-time weather data..."}}

chat(client, system_prompt, history, prompt)

@spec chat(client(), String.t(), list(), String.t()) :: response()

Sends a chat message with system prompt, history, and user prompt.

Parameters

  • client: The client instance
  • system_prompt: The system prompt to set context
  • history: List of previous messages in the conversation
  • prompt: The current user message

Examples

iex> client = AxiomAi.new(:vertex_ai, %{project_id: "my-project"})
iex> AxiomAi.chat(client, "You are a helpful assistant", [], "Hello!")
{:ok, %{response: "Hello! How can I help you today?"}}

complete(client, prompt, options \\ %{})

@spec complete(client(), String.t(), map()) :: response()

Generates completions based on a prompt.

Parameters

  • client: The client instance
  • prompt: The prompt to complete
  • options: Additional options (optional)

Examples

iex> client = AxiomAi.new(:vertex_ai, %{project_id: "my-project"})
iex> AxiomAi.complete(client, "The sky is", %{max_tokens: 10})
{:ok, %{completion: "blue and clear today."}}

embed(client, text, opts \\ %{})

@spec embed(client(), String.t(), map()) :: response()

Generates an embedding vector for a single text.

Options

  • :model - Embedding model (default depends on provider)
  • :task_type - e.g. "RETRIEVAL_DOCUMENT", "RETRIEVAL_QUERY"
  • :dimensions - Output dimensionality (default: 768)

Examples

iex> client = AxiomAi.new(:vertex_ai, %{project_id: "my-project"})
iex> {:ok, %{embedding: values}} = AxiomAi.embed(client, "What is AI?")
iex> length(values)
768

new(provider, config \\ %{})

@spec new(provider(), config()) :: client()

Creates a new client for the specified AI provider.

Parameters

  • provider: The AI provider to use (:vertex_ai, :openai, :anthropic, :local)
  • config: Provider-specific configuration map

Examples

iex> AxiomAi.new(:vertex_ai, %{project_id: "my-project", region: "us-central1"})
%AxiomAi.Client{provider: :vertex_ai, config: %{...}}

stream(client, message)

@spec stream(client(), message()) :: {:ok, Enumerable.t()} | {:error, any()}

Streams a chat message to the AI provider.

Parameters

  • client: The client instance
  • message: The message to send

Examples

iex> client = AxiomAi.new(:vertex_ai, %{project_id: "my-project"})
iex> {:ok, stream} = AxiomAi.stream(client, "Tell me a story")

# Process the stream
stream |> Enum.each(fn
  {:chunk, chunk} -> IO.write(chunk)
  {:status, code} -> IO.puts("Status: " <> inspect(code))
  {:error, reason} -> IO.puts("Error: " <> inspect(reason))
end)

stream(client, system_prompt, history, prompt)

@spec stream(client(), String.t(), list(), String.t()) ::
  {:ok, Enumerable.t()} | {:error, any()}

Streams a chat message with system prompt, history, and user prompt.

Parameters

  • client: The client instance
  • system_prompt: The system prompt to set context
  • history: List of previous messages in the conversation
  • prompt: The current user message

Examples

iex> client = AxiomAi.new(:vertex_ai, %{project_id: "my-project"})
iex> {:ok, stream} = AxiomAi.stream(client, "You are a helpful assistant", [], "Hello!")

# Process the stream
stream |> Enum.each(fn
  {:chunk, chunk} -> IO.write(chunk)
  {:status, code} -> IO.puts("Status: " <> inspect(code))
  {:error, reason} -> IO.puts("Error: " <> inspect(reason))
end)