Vibe Coding Discoverguidance
View on GitHubA guidance language for controlling large language models.
A Python framework for controlling LLM generation with composable prompts, conditionals, tool use, and grammar-based constraints. Supports structured outputs and multiple inference backends, including Transformers, llama.cpp, and OpenAI.
Use Cases
Constrain model output with regular expressions or grammarsGenerate schema-valid JSONCompose prompts with Python control flowBuild structured chat interactionsReduce generation steps through token fast-forwarding
Built With
- Language
- Jupyter Notebook
- Frameworks
- Transformers · llama.cpp · OpenAI API · Pydantic · Jinja2
Tags
LLM control · constrained generation · structured output · grammar · JSON schema · prompt programming · Python