What problem does it solve? LLMs often produce malformed JSON or free-form text that breaks downstream parsing. This Skill constrains token generation at the logit level so outputs always match a JSON schema, Pydantic model, regex, or fixed set of choices, eliminating retry loops and validation failures. ## Core Features & Use Cases - Guaranteed structured output: Compile Pydantic models, JSON schemas, regex patterns, or Literal choices into token-level automata that filter invalid tokens during generation. - Multiple local backends: Works with Transformers, llama.cpp, and vLLM via outlines.from_transformers, from_llamacpp, and from_vllm, plus OpenAI for server-side constrained JSON. - Use Case: Extract structured company information (name, founded year, industry, employees) from unstructured text by defining a Pydantic model, calling model(prompt, CompanyInfo), and validating the returned JSON string with CompanyInfo.model_validate_json(result). ## Quick Start Use the outlines skill to extract a user's name, age, and email from this text into a validated Pydantic model using a local Transformers model.
