For Engineers

Getting started with Adaline is simple:

• Sign up at adaline.ai

Upon registration, you’ll automatically receive:

• A default teamspace to organize your work.
• A project to contain your prompts.
• A prompt template to begin customizing.

1. In the sidebar, click Settings → Providers.
2. Click Add Provider and choose OpenAI.
3. Paste your OpenAI API key and save.

4. Your account is now linked—your connected models will appear in the editor.

1. Switch to Editor pillar by clicking on the Editor tab.
2. Click the model picker and choose a model (e.g., gpt-4oor o4-mini) from your connected provider.

1. In the Editor, update and edit the System and User messages to suit your use case.
2. Add or rename any variables.
3. Assign test values to each variable using the Variable Editor.

{{topic}}: "Science fiction novels"
{{tone}}: "Casual"

Build a robust test suite to evaluate your prompt:

1. Click on the Dataset tab and then click Add column and name the column exactly like the variable.

2. To organize the dataset within a project:

   • Click on the dataset name under your project and rename it.
   • Name your dataset descriptively.

3. Add 3-5 rows with different variable values to simulate various test cases. Bear in mind that each variable on the prompt template represents a single column. So if you have two variables, for example, topic and tone, then you essentially have two columns.

Create evaluation criteria to ensure consistent quality:

1. Switch to Evaluate Pillar by clicking on the Evaluate tab.

2. Click Add Evaluator. You get a list or the action menu of all Adaline’s available evaluation methods.

3. Select LLM as a Judge from the action menu and select your dataset that you’ve just created.

4. Write a clear evaluation rubric.

The response should be narrational and comprehensive.
The writing should be engaging and must maintain an imaginative style. 

Test your prompt against the dataset:

• Click Evaluate to run the evaluation.
• Review the results to identify any issues.
• Refine your prompt based on the evaluation feedback.

Make your prompt available for production use:

1. Once satisfied with the results, navigate to Deploy Pillar by clicking the Deploy tab.
2. Clickdeploy to make your prompt accessible via API.
3. Your prompt is now ready to be integrated into your application or workflow.