Enforce regex and grammar constraints during LLM token generation.
npx skills add https://github.com/graniet/kheish --skill guidance-graniet
Please help me install this Agent Skill. Skill: guidance Source: https://github.com/graniet/kheish/tree/main/skills/mlops/inference/guidance Command: npx skills add https://github.com/graniet/kheish --skill guidance-graniet
LLMs frequently generate malformed structured outputs such as invalid JSON, incorrectly formatted emails, or broken code when prompted for specific formats, requiring extensive post-processing, retry loops, and validation logic that slows development and introduces failure points in production workflows.
Use the guidance skill to extract a structured user profile with name, age, and email fields from the unstructured support ticket text "User Jane Smith, 28, [email protected] contacted support about a billing issue on 2024-10-05."
High-intent search queries and answers about installing and using this skill.
To guarantee valid JSON output from an LLM, apply regex and grammar constraints during token generation. This constrained generation approach eliminates malformed structured outputs and removes the need for post-processing or retry loops in production workflows.
Extract structured data from unstructured text by building multi-step agent workflows with Pythonic control flow. You can extract user information like names, emails, and dates directly into a guaranteed valid format suitable for database insertion.
Constrained generation is backend-agnostic and works with major LLM providers including OpenAI, Anthropic Claude, local Hugging Face Transformers models, and llama.cpp. This backend-agnostic integration allows flexible deployment across different environments without changing your workflow logic.
Token healing automatically fixes tokenization boundary issues during LLM inference to produce natural, correctly formatted outputs. It prevents awkward spacing or broken tokens that often occur when LLMs generate structured data like JSON or XML.
LLMs generate broken code and invalid XML because standard inference lacks format constraints during token generation. Applying grammar-based generation rules enforces syntax compliance during inference, eliminating invalid results without relying on post-generation validation logic.
automatedigitalEnforce regex and grammar constraints during LLM inference for valid structured outputs.
ovachieverConstrain LLM outputs to valid JSON, XML, and code using regex and grammar constraints.
zerwizConstrain LLM outputs with regex and grammar rules for structured formats.
cxnaiveExtract structured data from LLM responses using Pydantic schema validation.
