LangExtract practice with Ruby and Pycall.rb ( and uv and Ollama )

langextract.rb

require "pycall"

# see https://github.com/google/langextract/blob/main/examples/notebooks/romeo_juliet_extraction.ipynb

ENV["PYTHON"] = File.join(__dir__, ".venv/bin/python")
ENV["PYTHONPATH"] = File.join(__dir__, ".venv/lib/python3.14/site-packages")

pylx = PyCall.import_module "langextract"
pybuiltins = PyCall.import_module "builtins"

prompt = <<EOD
Extract characters, emotions, and relationships in order of appearance.
Use exact text for extractions. Do not paraphrase or overlap entities.
Provide meaningful attributes for each entity to add context.
EOD

examples = [
  pylx.data.ExampleData.new(
    text: "ROMEO. But soft! What light through yonder window breaks? It is the east, and Juliet is the sun.",
    extractions: [
      pylx.data.Extraction.new(
        extraction_class: "character",
        extraction_text: "ROMEO",
        attributes: {"emotional_state": "wonder"}
      ),
      pylx.data.Extraction.new(
        extraction_class: "emotion",
        extraction_text: "But soft!",
        attributes: {"feeling": "gentle awe"}
      ),
      pylx.data.Extraction.new(
        extraction_class: "relationship",
        extraction_text: "Juliet is the sun",
        attributes: {"type": "metaphor"}
      ),
    ]
  )
]

r = pylx.extract(
  text_or_documents: "Lady Juliet gazed longingly at the stars, her heart aching for Romeo",
  prompt_description: prompt,
  examples: examples,
  model_id: "gemma3:1b",
  model_url: "http://localhost:11434"
)

pp r.extractions.map { |e| pybuiltins.vars(e).to_h }

Ruby -> Python

Class()

を

Class.new

にすればイケそう。

Python -> Ruby

deep にすべての Object が map されるわけではなさそうなので、自前でどうにかする。上の例は

• AnnotatedDocument の extractions プロパティが Extraction の List になっている
• List は PyCall::List になって Ruby っぽく扱える
• Extraction は残念ながら Object になっているのでそのままでは便利に扱えないので vars() で Dict にしている
• vars() の時点でプロパティが Dict の key-value になって、Dict は PyCall::Dict に変換される
• PyCall::Dict にも Ruby らしい便利メソッドがあり、ここで to_h で Hash に持ってこれる