Enforce regex and grammar constraints on LLM token generation for structured outputs.
npx skills add https://github.com/ZardLi1115/zedclaw --skill guidance-zardli1115
Please help me install this Agent Skill. Skill: guidance Source: https://github.com/ZardLi1115/zedclaw/tree/main/optional-skills/mlops/guidance Command: npx skills add https://github.com/ZardLi1115/zedclaw --skill guidance-zardli1115
Guidance helps you stop LLM responses from producing malformed or unpredictable outputs by enforcing syntax, structure, and allowed tokens during generation.
Use the guidance skill to generate a valid JSON object by specifying a grammar that constrains each field (e.g., name, age, email) and then read the parsed fields directly from the result.
High-intent search queries and answers about installing and using this skill.
You can constrain LLM output to valid JSON by applying regex patterns or grammar constraints during generation, restricting allowed tokens to guarantee structural conformance and prevent malformed syntax.
Generating structured data like XML and code requires grammar-constrained generation, which enforces specific syntax rules and allowed tokens at generation time to ensure complex nested outputs remain perfectly valid.
Yes, constrained generation supports backend-flexible execution, allowing you to run regex and grammar constraints across OpenAI, Anthropic, Transformers, and llama.cpp backends within multi-step agent workflows.
You validate specific LLM output fields like emails and phone numbers using regex-based constrained generation, which forces the model to match required patterns and formatting during the generation process.
Token healing fixes spacing and boundary issues when combining prompt text with constrained generations, ensuring clean concatenation for structured outputs and multi-step agent workflows without breaking formatting.
LLMs produce unpredictable outputs in multi-step agent workflows because unconstrained generation allows arbitrary tokens; applying grammar constraints restricts allowed tokens to guarantee deterministic structure and valid syntax.
manutejAutomate constrained LLM generation with grammar-like constraints and typed placeholders.
devMoezConstrain LLM generation to produce valid JSON, XML, or code.
ovachieverConstrain LLM outputs to valid JSON, XML, and code using regex and grammar constraints.
musical-basicsConstrain LLM output with regex and grammar rules for valid JSON, XML, and code.
