GlueML is an innovative tool that simplifies the development and deployment of machine learning-powered APIs. At the core of this tool is the configuration format that defines the inputs, the outputs and the logic that needs to be exposed in those APIs.

This configuration file needs to be reachable by GlueML and use a very simple YAML format that defines the input and output of the JSON API that will be generated for your use case.

Take a look at the Hello World blog post for a simple introduction.

The GlueML YAML format is composed of the following sections:

version
This field must be 0.1 for the current version of GlueML

input
This section includes the list of parameters that your API will receive in the request.

output
This section includes the list of parameters that your API will return in the response.

prompt
Describe the logic to be executed by your API. it is described in natural language and used as a template including the input fields defined in the input section.

tools
The list of external tools to be used. For now only "web" tool is publicly available.

data (coming soon)
Data sources to be used while executing the logic of your API. For example this can be a link to a document or a S3 bucket with a list of files.

model
This field defines the type of model to be used to execute the prompt of your API. For free accounts the only available model is "free".