Skip to content

Latest commit

 

History

History
95 lines (84 loc) · 1.86 KB

dev.datamodel.md

File metadata and controls

95 lines (84 loc) · 1.86 KB

Overview

This documents describes the data schema used for running RAG experiments to evaluate system performance.

Data Schema

Datasources

Description

Datasources describe a collection of documents that can be used in a Test Run.

Schema

id name description

Documents

Description

A Document is a reference to the text content that will be used for generating Questions and performing the search.

Schema

id datasource_id name location

QASet

Description

A QASet is a reference to a location for a set of data composed of Questions and Correct Answers for a corresponding Document.

Schema

id datasource_id document_id name location

Question

Description

An Question and a corresponding Correct Answer based on information in a corresponding Document.

Schema

id qaset_id document_id question answer

TestRun

Description

A Test Run is a single test that was run on a set of Documents in a Datasource

Schema

id datasource_id description

Responses

Description

A Response is an answer to a corresponding Question, which was generated by your RAG system.

Scheam

id test_run_id question_id response

Context

Description

A Context is a portion of a Document that was returned with the Response by your RAG system.

Schema

id response_id text similarity_score

EvalFunction

Description

An Eval Function is a reference to a function that will be executed to evaluate the Response and Context to generate a score.

Schema

id name description function_index

TestEvalConfig

Description

A Test Eval Config is a set of Eval Function that will be run for a particular Test Run.

Schema

id test_run_id eval_function_id

ResponseEval

Description

A Response Eval is the score returned by an Eval Function for a given Reponse and Context

Schema

id test_run_id question_id response_id test_eval_config_id eval_score