Using Prompts in the Dashboard
Create and edit prompts, insert variables, test with values, and manage versions in the BrowserStack AI Evals dashboard.
Using Prompts in the Dashboard
The BrowserStack AI Evals dashboard provides a full-featured prompt editor with version history, variable detection, and a live testing playground.
Create a Prompt
Navigate to Prompts in the left sidebar.
Click New Prompt to open the prompt editor. or Select the dropdown to choose between Chat (default) and Text based prompt.
Enter a Name for your prompt. Names must be alphanumeric with hyphens and underscores only (e.g. summarizer, qa-assistant).
Write your prompt content in the editor. Click Save to create version 1.
Insert Variables
Variables in prompts use {{double_braces}} Mustache syntax. The dashboard detects variables automatically as you type.
In the prompt editor, type {{ to start a variable. The editor highlights detected variables as you write.
Give each variable a descriptive name, e.g. {{language}}, {{user_question}}, {{article}}. Variable names are case-sensitive.
For chat prompts, variables can appear in any message's content field — system messages, user turns, or assistant prefills.
The detected variables are listed in the Variables panel on the right. This panel updates live as you edit the prompt.
Test a Prompt with Variables
Use the Playground panel to test your prompt with real values before saving.
Open the Playground tab in the prompt editor (or click Test from the prompt list).
The Variables panel shows all detected {{placeholders}} from your template. Fill in a test value for each variable.
Click Run to compile the template and send it to the configured LLM. The compiled prompt and the model response appear in the output panel.
Adjust variable values and re-run to iterate. The prompt template is not modified by test runs.
Version History
Every save creates a new immutable version. The latest label always points to the most recent version.
Open a prompt and click the History tab.
The version list shows each version number, its labels, commit message, and creation date.
Click any version to view its content.
To compare two versions, open the version history tab and click Compare icon. A side-by-side diff highlights added, removed, and changed lines.
Manage Labels
Labels are human-readable aliases for specific versions (e.g. production, staging, v2). Your SDK fetches the version a label points to, so you can promote a new version without changing application code.
Open the prompt and go to the History tab.
Click the Labels button next to any version.
Add or remove labels. Multiple versions can share a label — the label always resolves to the version it was most recently assigned to.
Click Save Labels. The SDK cache is invalidated automatically so the next fetch returns the newly labelled version.
Some labels (e.g. latest) are protected and cannot be manually assigned — they are managed automatically by the platform.
Attach Tools to a Prompt
Open a prompt in the editor.
Scroll to the bottom section and click Tools.
Search for the tool by name and select the version or label to attach.
Save the prompt. When fetched via SDK, prompt.tools returns a ToolList that can be compiled and passed directly to your LLM API call.