> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ticketcord.com/llms.txt
> Use this file to discover all available pages before exploring further.
# The /train Command
> Teach your bot from historical ticket conversations
The `/train` command lets you teach your bot from past ticket conversations. It analyzes resolved tickets and extracts question-answer pairs that are added to your Knowledge Base.
The `/train` command requires a **Pro** or **Enterprise** plan.
## What It Does
The `/train` command:
1. Scans your resolved tickets from the past X days
2. Uses AI to identify question-answer pairs
3. Extracts the best Q\&A from each conversation
4. Adds them to your Knowledge Base automatically
This captures the "tribal knowledge" your team has built up - answers to common questions that may not be in your documentation.
## Running /train
### Basic Usage
In any channel where your bot is active, run:
```
/train
```
The bot will analyze tickets from the last 90 days (default).
### Specifying Days
To analyze a different time period:
```
/train days:30
```
This analyzes tickets from the last 30 days.
| Parameter | Range | Default |
| --------- | ----- | ------- |
| days | 7-365 | 90 |
Start with 30-60 days to test. Increase to 90+ once you're confident it's working well.
### Requirements
To run `/train`, you need:
* **Administrator** permission in the server
* Knowledge Base **enabled** in your dashboard
* At least one **resolved** ticket in the time period
## The Training Process
After running `/train`, you'll see:
The bot counts how many tickets are available to analyze.
AI reads through each ticket conversation, looking for:
* Clear questions from users
* Helpful answers from staff
* Complete resolution patterns
The AI extracts question-answer pairs that would help future users.
Extracted Q\&A pairs are stored and used for future suggestions.
### Progress Updates
While training, you'll see updates like:
```
🤖 Training in Progress
Analyzing stored tickets from the last 90 days...
📊 Progress: 15/42 tickets (35%)
✅ Extracted: 8 Q&A pairs
[Cancel Training]
```
### Cancelling Training
Click the **Cancel Training** button to stop training early. Any Q\&A pairs already extracted will be kept.
## First-Time Setup
### AI Consent
The first time you run `/train`, you'll need to accept AI data processing terms.
**Why is this required?**
* Ticket conversations are sent to AI for analysis
* This may include user messages and staff responses
* You're authorizing this processing
Only the **bot owner** can accept AI consent terms. If you're not the owner, ask them to run `/train` first.
### Consent Prompt
You'll see:
```
📋 AI Data Processing Consent Required
To use /train, we need your consent to process ticket conversations with AI.
What happens:
• Ticket messages are analyzed to extract Q&A pairs
• AI learns from staff responses to help future users
• Data is processed securely and not shared
[Accept] [Decline]
```
Click **Accept** to proceed, or **Decline** to cancel.
## What Makes Good Training Data
The AI looks for tickets that have:
### Clear Questions
```
User: How do I cancel my subscription?
```
### Helpful Answers
```
Staff: Go to Settings → Subscription → Cancel.
Your access continues until the billing period ends.
```
### Complete Resolution
Tickets that were marked as resolved or closed with a solution.
## What Gets Skipped
The AI skips tickets that:
* Were closed without a real solution
* Had mostly back-and-forth troubleshooting
* Contained only small talk or greetings
* Were spam or test tickets
## After Training
### Viewing Results
After training completes, you'll see:
```
✅ Training Complete!
📊 Results:
• Tickets analyzed: 42
• Q&A pairs extracted: 12
• Tokens used: 150K
Your bot is now smarter! New Q&A will be used in future suggestions.
```
### Using Trained Knowledge
Trained Q\&A pairs are automatically used when:
* Users open new tickets with similar questions
* The Knowledge Base searches for answers
* No better documentation source exists
### Retraining
You can run `/train` again to:
* Capture newer ticket conversations
* Learn from recently resolved issues
* Update knowledge after your team's processes change
There's a cooldown between training runs to prevent overuse. You'll see how long to wait if you try to train too soon.
## Best Practices
### Train After Major Updates
Run `/train` after:
* Launching new features
* Resolving new types of issues
* Your team learns new troubleshooting steps
### Combine with Documentation
`/train` works best alongside documentation sources:
* Documentation provides official information
* Trained Q\&A provides real-world answers
* Together they cover more questions
### Regular Training Schedule
Consider training:
* Monthly for active support teams
* After resolving 50+ new tickets
* When you notice AI missing common questions
### Review the Results
Check your Knowledge Base after training:
* Look at the extracted Q\&A pairs
* Verify they're accurate
* Delete any that aren't helpful
## Troubleshooting
### "No Tickets Found"
This means:
* No resolved tickets in the time period
* Tickets don't have enough content
* Knowledge Base isn't enabled
**Solution:** Check that you have resolved tickets and Knowledge Base is enabled.
### "Training Stalled"
If training shows no progress for 5+ minutes:
* The process may have encountered an error
* Try cancelling and running again
* Check for bot connectivity issues
### "Consent Required" for Non-Owner
Only the bot owner can accept AI consent.
**Solution:** Ask the bot owner to run `/train` first and accept consent.
### Low Q\&A Extraction
If few Q\&A pairs are extracted:
* Tickets may not have clear question-answer patterns
* Staff responses may be too brief
* Try a longer time period (more tickets)
Get instant support via private DM