You will see multiple templates with information about each template. To see more details about a template, click on the Preview button. This will open a dialog box with options to view Golden QnA, Datasets, Retrieval settings, and Generation settings.
Look for the template named “Chat over text files” and click on the Use Template button. This will open the Create LLM App tab with configured hyperparameters, Golden QnA, and other settings.
Click on the Create button in the top right corner. This will create the application with the selected settings and open the Experiments tab. In this tab, you will see the application listed. Click on the Select button to select this application and proceed to the next phase.
After selecting the application, you will see multiple combinations/configurations of your application. Click on the Initialize Combinations button to initialize these combinations. Initialization involves preparing uploaded data, chunking, creating embeddings, and uploading to the vector database.
Once initialization is complete, the combinations are ready to execute. To execute combinations, first, select them by checking the checkbox in the leftmost column. Then, click on the Run Selected button.
Once each combination is completed, you can view the accuracy, latency, and cost of each combination. By clicking on the View button, you can see the responses generated against Golden QnA, evaluations, and tracing.
In the view dialog box, you can also run other questions with these settings for testing. Additionally, you can download the evaluation results for all questions.
After analyzing each combination, select the one that is most efficient in terms of accuracy, cost, and latency. Click on the Select button in the status column (the rightmost column) to choose the combination. Then, proceed to the Generation Optimization tab.
In this tab, you can see both the Retrieval and Generation combinations that you selected. Further test these combinations together by clicking on the Test button. This will open a chat window where you can test with additional questions.
Once you are satisfied, click on the Deploy button to deploy the application. After deployment, click on the Generate API Key button to generate an API key for this deployed application.
You can learn how to use this API key by clicking on the < > button in the Generate API Key tab.