llm-structured-output

What problem does it solve?

Replaces ad-hoc parsing of LLM responses with typed, validated data by extracting JSON-like structures from model outputs.

Core Features & Use Cases

• Structured extraction across OpenAI, Anthropic, and Google Gemini APIs using provider-specific strategies (response_format with json_schema, tool_use blocks, and responseSchema) to guarantee schema conformance.
• Schema-driven validation and retry logic to handle decoding failures, malformed JSON, or missing fields in production systems.
• Supports defining target schemas for downstream pipelines, databases, or UI layers, with step-by-step guidance on choosing the right method per provider.

Quick Start

Create a concrete target schema, configure the extraction provider, and run a single-pass extraction on your LLM output to obtain a typed data object.