Test runner

Overview

The Test Runner lets you paste a real customer message and see exactly how the pipeline would process it. This is the fastest way to verify your configuration (routing rules, scoring weights, templates, Signal Guard, and brand voice) without sending real messages.

How to use

1. Enter a message: paste or type a customer message in the text area.
2. Set contact details (optional): name, phone, email. These populate template variables.
3. Choose a channel: SMS or email. This affects which templates match and reply length constraints.
4. Click "Run Test": the pipeline processes the message and displays results.

Reading results

Results are displayed in pipeline order:

Classification (requires AI)

Shows the AI's analysis of the message:

• Intent: what the customer wants (for example, booking_request, complaint, general_inquiry)
• Urgency: how time-sensitive: low, medium, high, critical
• Sentiment: emotional tone: positive, neutral, negative
• Confidence: how certain the AI is (0–1)
• Summary: brief AI-generated description
• Extracted Data: structured fields pulled from the message (client name, deadlines, etc.)

If no AI model is configured, classification is skipped and a note explains how to enable it.

Lead score

Shows the composite score and tier:

• Score: numeric value (0–100) with a color-coded progress bar
• Tier: hot (red), warm (orange), cool (blue), cold (gray)
• Recommended Actions: what the scoring engine suggests based on the tier

Matched rules

Lists every routing rule that matched the test signal:

• Rule name with a green checkmark
• Actions that would fire, shown as badges (for example, send_template, create_task, escalate)

If no rules match, you'll see "No routing rules matched this signal." This means you need to add or adjust rules.

Generated reply

If AI reply generation is enabled and a reply was generated:

• Channel: SMS or email
• Requires approval: yellow warning if the reply needs human review before sending
• Reply body: the actual text that would be sent

Actions triggered

Summary of all actions that would execute, with their status (success/dry-run).

Pipeline errors

If any stage fails (for example, classification timeout, scoring error), errors are shown with the stage name and error message. The pipeline continues past failures. A classification error doesn't block routing.

What the test runner does not do

• Does not send real messages. No SMS, email, or webhook is actually sent.
• Does not update CRM. update_crm actions run in dry-run mode.
• Does not create real tasks. Tasks are simulated.
• Does not start background jobs. Nurture sequences and SLA monitors are not started.
• Does not count toward usage. Test runs are free and don't affect your signal quota.

Example test messages

Try these to verify different parts of your configuration:

Missed call scenario (zero AI):

Hi, I missed your call. Can you call me back?

Hot lead with budget:

I need a kitchen remodel done by March. Budget is $40k. I'm the homeowner and ready to sign a contract this week.

After-hours inquiry:

(Run this outside your configured business hours) Do you have any availability this week for a consultation?

Compliance-sensitive (Signal Guard):

What medication should I take for my skin condition? I've been having allergic reactions.

Booking request:

I'd like to schedule an appointment for next Tuesday at 2pm if possible.

Tips

• Test after every configuration change. Changed a routing rule? Run a test. Adjusted scoring weights? Run a test.
• Use real customer messages. Copy-paste actual messages from your inbox for the most realistic results.
• Test edge cases. Try after-hours messages, messages with no clear intent, messages in different languages.
• Check "No AI" behavior. If you're not sure whether your rules work without AI, test with a message and see if rules still match on channel/source/business hours conditions.
• Compare before and after. Run the same test message before and after changing scoring weights to see the impact.