outlines

Outlines is a high-performance library for guided and structured text generation with LLMs, developed by the .txt (dottxt-ai) team. It guarantees that a model's output strictly adheres to a predefined format—such as JSON schemas, regular expressions, specific types, or context-free grammars (CFG)—directly during token generation by dynamically masking invalid tokens at the logit level. By avoiding expensive post-parsing and retry loops, Outlines provides zero-overhead, highly reliable structured outputs. It is fully compatible with major API providers and open-source inference engines, including OpenAI, vLLM, Ollama, llama.cpp, and Hugging Face Transformers.

• 100% guaranteed structured output via FSM token masking
• Zero inference latency overhead and zero retry loops
• Universal backend compatibility (OpenAI, vLLM, Ollama, Transformers, etc.)
• Type-driven interface with native Pydantic, Regex, and CFG support