Skip to content

README.md

Latest commit

 

History

History
40 lines (24 loc) · 1.99 KB

README.md

File metadata and controls

40 lines (24 loc) · 1.99 KB
Raw

Generative Regression Analysis Framework for Issue Tracking and Evaluation - Running Client Locally

Installation

To set up and run the application for local development, follow these steps after configuring the server and database:

npm i
npm run dev

Before committing, run npm run fix to apply quick styling corrections. For other linting scripts, please check package.json

Database

The current app uses MongoDB as the database. For detailed information on setting up and managing the database, please refer to the server README.

Model Endpoints

There are two places in the app where the model endpoints are used. To enable these features, ensure that the appropriate environment variables are set.

  1. QA Test Creation: We support Ollama for model inferencing to create QA tests
  2. Test Run Pipeline: Although test run pipeline is handled by the server, in case you use WatsonX AI, you can select the model to evaluate from GUI dropdown by setting up the environment variables.

The models used by the application can be specified through the following environment variables:

  • NEXT_PUBLIC_JUDGE_MODEL: Model used for the LLM judge validation. Default model is llama3.3 (llama3.3:70b).
  • NEXT_PUBLIC_DESIRED_OUTPUT_MODEL: Model for desired ouptut. Default model is mistral-small3.1 (mistral-small3.1:24b).
  • NEXT_PUBLIC_TEST_DEFAULT_MODEL: Model for regression tests issue validation. Default model is granite4:small-h
  • NEXT_PUBLIC_TEST_DEFAULT_MODEL_DISPLAY: The same model as the above but a name to be displayed on the UI. Default value is granite4:small-h

If you use Ollama for inference, please make sure that you have the model available on your machine.

Constant Values

Default inference parameters are on utils/constants.ts

Auth

By default, the app bypasses the authorization configured through NextAuth. To enable authorization, modify auth.ts file.