instructor_ex

Structured, Ecto outputs with OpenAI (and OSS LLMs)

Instructor.ex is a spiritual port of the great Instructor Python Library by @jxnlco, check out his talk on youtube. This library brings structured prompting to LLMs. Instead of receiving text as output, Instructor will coax the LLM to output valid JSON that maps directly to the provided Ecto schema. If the LLM fails to do so, or provides values that do not pass your validations, it will provide you utilities to automatically retry with the LLM to correct errors. By default it's designed to be used with the OpenAI API, however it provides an extendable adapter behavior to work with ggerganov/llama.cpp and Bumblebee (Coming Soon!).

At its simplest, usage is pretty straightforward,

defmodule SpamPredicition do
  use Ecto.Schema
  use Instructor.Validator

  @doc """
  ## Field Descriptions:
  - class: Whether or not the email is spam
  - reason: A short, less than 10 word rationalization for the classification
  - score: A confidence score between 0.0 and 1.0 for the classification
  """
  @primary_key false
  embedded_schema do
    field(:class, Ecto.Enum, values: [:spam, :not_spam])
    field(:reason, :string)
    field(:score, :float)
  end

  @impl true
  def validate_changeset(changeset) do
    changeset
    |> Ecto.Changeset.validate_number(:score,
      greater_than_or_equal_to: 0.0,
      less_than_or_equal_to: 1.0
    )
  end
end

is_spam? = fn text ->
  Instructor.chat_completion(
    model: "gpt-3.5-turbo",
    response_model: SpamPredicition,
    max_retries: 3,
    messages: [
      %{
        role: "user",
        content: """
        You purpose is to classify customer support emails as either spam or not.
        This is for a clothing retailer business.
        They sell all types of clothing.

        Classify the following email: #{text}
        """
      }
    ]
  )
end

is_spam?.("Hello I am a Nigerian prince and I would like to send you money")

# => {:ok, %SpamPredicition{class: :spam, reason: "Nigerian prince email scam", score: 0.98}}

Simply create an ecto schema, optionally provide a @doc to the schema definition which we pass down to the LLM, then make a call to Instructor.chat_completion/1 with context about the task you'd like the LLM to complete. You can also provide a validate_changeset/1 function via the use Instructor.Validator which will provide a set of code level ecto changeset validations. You can use this in conjunction with max_retries: 3 to automatically, iteratively go back and forth with the LLM up to n times with any validation errors so that it has a chance to fix them.

Curious to learn more? Unsure of how you'd use this? Check out our extensive set of tutorials

Configuration

To configure the default OpenAI adapter you can set the configuration,

config :openai, api_key: "sk-........"
config :openai, http_options: [recv_timeout: 10 * 60 * 1000]

To use a local LLM, you can install and run llama.cpp serer and tell instructor to use it,

config :instructor, adapter: Instructor.Adapters.Llamacpp

Installation

In your mix.exs,

def deps do
  [
    {:instructor, "~> 0.0.2"}
  ]
end

TODO

Blog Posts

Why structured prompting?

Meditations on new HCI. Finally we have software that can understand text. f(text) -> text. This is great, as it gives us a new domain, but the range is still text. While we can use string interpolation to map Software 1.0 into f(text), the outputs are not interoperable with Software 1.0. Hence why UXs available to us are things like Chatbots as our users have to interpret the output.

Instructor, structure prompting, gives use f(text) -> ecto_schema. Schemas are the lingua franca of Software 1.0. With Instrutor we can now seamlessly move back and forth between Software 1.0 and Software 2.0.

Now we can maximally leverage AI...
From GPT-4 to zero-cost production - Distilation, local-llms, and the cost structure of AI.

... 😘

brianpetro/instructor_ex

instructor_ex

Configuration

Installation

TODO

Blog Posts