What problem does it solve? LLMs often produce malformed JSON or free-form text that breaks downstream parsing. This Skill guarantees structurally valid outputs by constraining token generation at the logit level against JSON schemas, Pydantic models, regex patterns, or literal choices. ## Core Features & Use Cases - Guaranteed Valid Structure: Compile output types into token-level automata so invalid JSON, enums, or regex-mismatched text cannot be generated. - Multiple Backends: Works with Transformers, llama.cpp, vLLM, and OpenAI via outlines.from_transformers, from_llamacpp, from_vllm, and from_openai factories. - Pydantic Integration: Pass a Pydantic model as the output type and validate the returned JSON string with model_validate_json. - Use Case: Extract structured company information (name, founded year, industry, employees) from unstructured text using a local Phi-3 or Llama model, with every field guaranteed to match your schema constraints. ## Quick Start Use the outlines skill to extract a user's name, age, and email from this text into a validated Pydantic model with a local Transformers model.
