Explore Mode
Ad-hoc quick testing without persistence
What Is Explore Mode?
Explore mode lets you run ad-hoc tests without creating a suite or persisting results. It is useful for quick iteration: change a variable, tweak an assertion, and see the result immediately.
Using Explore
In the Evals tab, switch to Explore. Provide:
- A prompt version to test against
- Input variables (as JSON)
- Optional model configuration
- Render and/or LLM assertions
Click Run to get instant results. Nothing is saved to the database.
Explore API
http
POST /api/prompts/{promptId}/eval/explore
{
"versionNo": 3,
"variables": { "name": "Alice", "tone": "friendly" },
"modelConfig": { "model": "gpt-4o-mini", "temperature": 0 },
"expectRender": [
{ "contains": "Alice" },
{ "word_count": { "min": 10 } }
],
"expectLlm": [
{ "sentiment": "positive" },
{ "latency": { "max": 2000 } }
]
}The response includes the rendered output, LLM response (if applicable), assertion results, and an overall status (passed/failed/errored).
When to Use Explore vs. Suites
| Use Explore When | Use Suites When |
|---|---|
| Trying a new assertion before adding it to a suite | You need persistent, trackable results |
| Debugging a single test case | You want to compare runs over time |
| Quick sanity check during development | You need quality gates for deploys |