Cloud-powered AI helpers backed by anu_cloud.
These primitives let you classify, extract, draft, and summarize WhatsApp
conversation text without managing your own LLM provider. Pair them with
Anu.Message to build conversational flows in a few lines.
Client
Every function takes an %Anu.Client{} as its last argument. The client
carries the cloud API key, cloud URL and Finch instance:
client =
Anu.Client.new(
finch: MyApp.Finch,
cloud_api_key: System.fetch_env!("ANU_API_KEY")
)Example
with {:ok, %{"intent" => "support"}} <-
Anu.AI.classify(text, [intents: ["support", "sales", "spam"]], client),
{:ok, %{"reply" => reply}} <-
Anu.AI.reply(text, [tone: "warm"], client) do
from
|> Anu.Message.new()
|> Anu.Message.text(reply)
|> Anu.deliver(client)
endAll functions return {:ok, decoded_map} or {:error, %Anu.Error{} | :cloud_not_configured}.
Response keys are strings (decoded JSON).
Summary
Functions
Classifies text against a list of intents. Returns {intent, confidence}.
Extracts structured data from text matching the given JSON schema.
Drafts a reply to the user text.
Summarizes a list of conversation messages into a short summary.
Types
@type result() :: {:ok, map()} | {:error, Anu.Error.t() | :cloud_not_configured}
Functions
@spec classify(String.t(), keyword(), Anu.Client.t()) :: result()
Classifies text against a list of intents. Returns {intent, confidence}.
Options
:intents(required) - list of candidate intent strings.
Examples
Anu.AI.classify("I need to reset my password", [intents: ["support", "sales", "spam"]], client)
#=> {:ok, %{"intent" => "support", "confidence" => 0.93}}@spec extract(String.t(), keyword(), Anu.Client.t()) :: result()
Extracts structured data from text matching the given JSON schema.
Options
:schema(required) - a map describing the fields to extract.
Examples
Anu.AI.extract("My name is Ana and my order is #123", [schema: %{name: "string", order_id: "string"}], client)
#=> {:ok, %{"data" => %{"name" => "Ana", "order_id" => "123"}}}@spec reply(String.t(), keyword(), Anu.Client.t()) :: result()
Drafts a reply to the user text.
Options
:context- extra background to ground the reply (string):tone- e.g."warm","professional","playful"
Examples
Anu.AI.reply("when does my package arrive?", [tone: "warm"], client)
#=> {:ok, %{"reply" => "Hi! Your package is on its way..."}}@spec summarize([map()], Anu.Client.t()) :: result()
Summarizes a list of conversation messages into a short summary.
Examples
Anu.AI.summarize([
%{role: "user", content: "hi"},
%{role: "assistant", content: "hello, how can I help?"}
], client)
#=> {:ok, %{"summary" => "User said hi; we offered help."}}