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 using finite state machines, so outputs always match your JSON schema, Pydantic model, or regex pattern without retry loops. ## Core Features & Use Cases - Schema-Guaranteed Generation: Convert Pydantic models or JSON schemas into grammars that filter invalid tokens during decoding, guaranteeing valid structured output. - Multiple Local Backends: Run constrained generation on Transformers, llama.cpp, or vLLM models, with multi-GPU and quantization support for production throughput. - Typed Generators: Use built-in generators for choices, integers, floats, regex patterns, and full JSON objects with nested models, enums, and field constraints. - Use Case: Extract structured company information (name, founded year, industry, employees) from hundreds of unstructured text snippets into validated Pydantic objects for a data pipeline. ## Quick Start Ask the AI to use Outlines with a Pydantic model to extract structured fields from your text and return a validated JSON object.
