Testing Your Agent¶
SpeakNode provides a built-in testing interface so you can interact with your agent directly in the browser before deploying it to phone lines or your website.
Opening the Test Page¶
There are two ways to access the test interface:
From the Agent Editor¶
- Open the agent you want to test.
- Click the Test button in the top-right corner of the editor.
- The test page opens in a new tab.
Note
The test page uses the latest saved configuration, including unpublished draft changes. This lets you iterate on your agent without publishing each time.
Public Test URL¶
For published agents, SpeakNode provides a public test URL that anyone can use.
- Open the agent and ensure it is published.
- Copy the Public Test URL from the agent settings.
- Share this URL with testers or stakeholders.
Warning
The public test URL uses the published version of the agent, not the draft. Make sure to publish your changes before sharing.
Test Page Layout¶
The test page is divided into two panels:
Transcript Panel¶
The left panel shows the conversation transcript in real-time:
- Agent messages appear on one side, user messages on the other.
- Each message is timestamped.
- New messages appear automatically as the conversation progresses.
Controls Panel¶
The right panel contains session controls:
- Start / Stop Session button to begin or end the conversation.
- Language Selector dropdown to choose which language to use for the session.
- Microphone Controls to mute/unmute your microphone.
Running a Test Session¶
Step 1: Select a Language¶
If your agent supports multiple languages, choose the desired language from the Language dropdown before starting the session.
Step 2: Start the Session¶
- Click the Start Session button.
- Grant microphone access when prompted by your browser.
- The agent will speak its first message (the greeting configured for the selected language).
Step 3: Have a Conversation¶
- Speak naturally into your microphone.
- Watch the transcript update in real-time as both you and the agent speak.
- The agent responds based on its system prompt and configuration.
Step 4: Stop the Session¶
Click the Stop Session button to end the conversation. The full transcript remains visible for review.
Testing tips
- Test all the edge cases described in your system prompt.
- Try interrupting the agent mid-sentence to verify interruption handling.
- Test each language separately.
- Verify that tools trigger correctly by speaking the expected phrases.
- Test with background noise to check VAD sensitivity.
Viewing the Transcript¶
After the session ends, the transcript is saved and can be reviewed:
- Scroll through the entire conversation in the transcript panel.
- Each turn shows the speaker and the transcribed text.
- If the session was linked to a conversation record, you can find it in the Conversations page for deeper analysis.
Troubleshooting¶
| Issue | Solution |
|---|---|
| No audio from agent | Check your browser's audio output settings and volume. |
| Microphone not detected | Ensure you granted microphone permission. Try a different browser. |
| Agent not responding | Verify the agent has at least one language configured with a first message. |
| Wrong language used | Select the correct language from the dropdown before starting the session. |