All Tools
G
LLMFreeOpen Source
GUIDANCE
A guidance language for controlling LLMs with constrained generation
MIT
ABOUT
Traditional LLM prompting lacks precise control over output structure and format, leading to unreliable results, high latency, and increased token costs from repetitive prompting. Guidance solves this by providing a programming paradigm where developers constrain LLM output with grammars, regex, and schemas while interleaving generation with control flow logic — guaranteeing structured output and reducing overhead.
INSTALL
pip install guidanceINTEGRATION GUIDE
1. Constrain LLM outputs with regex and context-free grammars for reliable structured generation
2. Generate valid JSON from Pydantic schemas for data extraction and API integration
3. Build multi-step prompting workflows with conditionals and loops for complex tasks
4. Implement tool use and function calling with guaranteed syntax compliance
5. Accelerate token generation through fast-forwarding cached prefix patterns
TAGS
llmconstrained-generationstructured-outputprompt-engineeringmicrosoftjson-generationnlplanguage-model