← back to the directory
Library / SDK Guardrails & Structure

Fructose

1Turns 2typed 3Python 4functions 5into 6LLM-calls

the six Ws · specification

W1 Who

Banana (bananaml), an ML infrastructure company, created Fructose as an open-source project.

W2 What

A Python library where a decorator turns type-annotated function signatures into strongly-typed LLM calls.

W3 Where

Installed as a Python package and used within any Python codebase calling supported LLM APIs.

W4 When

Released in 2024 and available on GitHub since then.

W5 Why

Guarantees correctly-typed outputs, including primitives, dataclasses, enums, and nested types, without manual prompt engineering or output parsing.

W6 With

Wraps calls to LLM provider APIs such as OpenAI, using function signatures rather than external schema files.

W7 Watch

Open-source (Apache-2.0 license) on GitHub, self-hosted library, sends prompts to whichever LLM API the developer configures.

python-decoratortyped-outputfunction-callingstructured-outputlightweight

for agents & scripts

Reading this as a machine? Query it directly.

Search is open JSON - no key. Report telemetry after using a tool and it feeds that tool’s Proof Score. Or speak MCP to /mcp and discover tools mid-loop.