Documentation
flowschema docs
How to check tool calls and structured answers against a JSON Schema with flowschema: the install, the two loops of an agent, provider strict modes, cut-off and streaming calls, and what is checked.
Start
From the install to a first check, and the two places an agent checks JSON.
- Getting startedflowschema is a Rust crate that checks JSON against a JSON Schema: a tool call against the tool's parameters, a structured answer against its shape. This page takes it from the install to a first…
- Tool calls and answersAn agent meets JSON Schema in two places: a tool's parameters, which every call the model writes must fit before the tool runs, and the shape of a structured answer, which the program reads back. In…
With a model
What a provider holds a model to, and what a model writes that is not yet a value.
- Provider strict modesA provider's strict mode constrains what the model can write, to a part of JSON Schema that differs from one provider to the next. This page says what each holds the model to, what is still the…
- Cut-off and streaming callsWhat a model writes does not always arrive whole or as the schema spells it: a reply stops at the output limit, a call's arguments come in pieces, a number comes as a string. This page covers the…
Reference
Every item of the API, what is checked and how it is held to it, and how fast.
- APIOne type, Schema, compiled from a JSON Schema and checked against once or many times, and two functions beside it: fit, for a provider's strict mode, and whole, for a text the output limit cut off.
- What is checkedflowschema checks JSON Schema as its drafts say: every keyword of validation, every applicator, references and formats. This page lists them, says what is left unchecked and how a program learns of…
- Benchmarksflowschema is measured against the jsonschema crate on the checks an agent makes, on real LLM schemas and answers, and for what it adds to a program. Every figure here is the latest run's in the…