Guidance
Microsoft's programming paradigm for controlling LLM generation with constrained outputs
Guidance
Microsoft's programming paradigm for controlling LLM generation with constrained outputs
Guidance is a programming paradigm from Microsoft Research that allows developers to constrain LLM generation with structured templates that interleave text, conditional logic, and generation calls, enabling deterministic control over output format while allowing the model to fill in the creative content. Its token healing and grammar-constrained generation capabilities ensure outputs conform to required formats without post-processing, making it particularly valuable for code generation and structured data extraction. Researchers and developers building AI applications requiring precise output control use Guidance to combine the flexibility of LLMs with the reliability of rule-based systems.
Key Features
- ✓Constrained generation
- ✓Structured templates
- ✓Grammar control
- ✓Token healing
- ✓Multi-model
Quick Info
- Category
- AI Developer Tools
- Pricing
- Free
More AI Developer Tools Tools
Langchain Hub
AI Developer ToolsCommunity hub for discovering, sharing, and managing reusable LangChain prompts, chains, and agents across…
Klu AI
AI Developer ToolsAI application development platform for building, testing, and deploying LLM-powered apps with prompt manag…
Outlines
AI Developer ToolsPython library for structured generation with LLMs that constrains model outputs to follow specific formats…
Phidata
AI Developer ToolsPython library and platform for building multimodal AI agents with memory, knowledge, and tools, with built…