Structured Output

Output format is a core challenge of integrating language models into larger workflows, and numerous libraries have been created to help.

• OpenAI Functions
• jxnl/instructor - function call wrapper with Pydantic model validation
• Semantic Kernal
• Guidance
• Jsonformer
• Kor - schema-driven output extraction
• Guardrails
• TypeChat - get well-typed responses from language models
• Marvin AI - structured output and other agent-building shortcuts
• Fructose
• GGML BNF - Constrain llama.cpp model inference with a formal grammar