Extract validated structured data from LLM responses into Pydantic models.
npx skills add https://github.com/MarbleSodas/Mavis --skill instructor-marblesodas
Please help me install this Agent Skill. Skill: instructor Source: https://github.com/MarbleSodas/Mavis/tree/main/optional-skills/mlops/instructor Command: npx skills add https://github.com/MarbleSodas/Mavis --skill instructor-marblesodas
Extracting structured fields from LLM outputs is often brittle: JSON can be malformed, required fields can be missing, and types don’t reliably match what your app expects.
Use instructor to extract a Pydantic model from a messy LLM response, automatically retrying until the output matches your schema.
High-intent search queries and answers about installing and using this skill.
Structured data extraction from LLM responses into Pydantic models is achieved by validating schemas, enforcing types and constraints, and feeding validation errors back to the model for automatic retries until the output matches.
JSON parsing fails when extracting nested fields because LLM outputs often contain malformed JSON, missing required fields, or mismatched types that do not align with your application's expected schema constraints.
Streaming partial Pydantic model results from OpenAI and Anthropic responses is supported, allowing you to process incremental structured outputs in real time for faster user experiences and incremental data pipelines.
Automatically retrying LLM data extraction on validation failure is handled by feeding Pydantic validation errors back to the model and re-attempting extraction until it succeeds, with configurable retry limits.
Structured output validation works with both OpenAI and Anthropic through provider-specific integrations, ensuring consistent response handling and Pydantic model enforcement across different LLM providers.
Enforcing enums and constraints when parsing complex JSON from LLM text is done by defining Pydantic models with specific types and rules, which the extraction process validates against before accepting the output.
ovachieverExtract structured data from LLM responses with Pydantic validation and retries.
JamesFincherConvert LLM responses into Pydantic-validated JSON with automatic retries.
fjacquetValidate FinWiz outputs against Pydantic v2 strict schemas.
MinhHoangDonoIntegrate Anthropic, OpenAI, and Gemini APIs for chat, tool use, and structured outputs.
