Beyond your website, there are three more ways to feed knowledge to your AI agent. Each serves different purposes and helps your bot handle specific scenarios your website content might miss.
What Data Sources Actually Do
Data Sources are pure knowledge - they dont control how your bot talks or behaves. That's handled by your system prompt in Settings.
Think of it this way:
Data Sources = What your agent knows
System Prompt = How your agent uses that knowledge
Now let's dive into the three additional ways to make your bot smarter.
Files - Upload Your Documents
Perfect for detailed info that lives in PDFs or documents outside your website.
What to Upload
Product catalogs with full specifications
Sizing charts and measurement guides
Return and exchange policies
Shipping information and zone details
Brand voice guidelines
FAQ documents from your internal team
Training manuals for complex products
Warranty information
How to Upload
Just drag and drop files into the Files section, or click to browse and select them.
File requirements:
PDF format is needed
Keep files under 100MB for smooth processing
Text-heavy PDFs train better than image-heavy ones
Scanned documents might not work as well as native PDFs
When Files Are Most Useful
You have detailed product specs not on your website
Your return policy is buried deep in legal pages
You want to train on internal documentation
You need the bot to know technical details that aren't customer-facing
Pro tip: If your website doesn't have comprehensive product details, upload your internal product catalogs. The bot can then answer detailed questions your site might not cover.
Text - Quick Knowledge Updates
This is your go-to for temporary information or quick updates that don't need a whole document.
Perfect Use Cases
Temporary closures: "We're closed Dec 24-26 for holidays"
Shipping delays: "All orders delayed 2 days due to weather"
Limited-time info: "Flash sale ends midnight Dec 27th"
Store hours and locations
Special instructions during busy periods
Staff availability: "Customer service available 9am-6pm EST"
How to Add Text Data
Click in the Text section
Type your information directly
Hit Save
Update or delete when the info changes
Text Data Best Practices
Keep it specific: Instead of "We might be slow," write "Response times may be 2-3 hours longer during Black Friday week."
Make it actionable: "For urgent issues, call 986-555-0123" gives customers an alternative.
Date your info: "Updated Dec 15: Holiday shipping cutoff is Dec 20th for standard delivery."
Q&A Pairs - Exact Answers for Common Questions
When customers ask the same question 50 different ways but the answer never changes, Q&A pairs are your best friend.
Setting Up Q&A Pairs Manually
Click Add New Q&A
Enter the question: "What fabric is the blue hoodie made from?"
Enter the exact answer: "100% organic cotton with pre-shrunk treatment"
Save it
The Smart Way - Improve Answers from Real Chats
When reviewing customer conversations, you might see the bot gave a response that wasn't quite right. Instead of letting it happen again:
Click the Improve Answer button on that message
Type what the bot should have said instead
This automatically creates a new Q&A pair
The customer's original question becomes the Q
Your corrected response becomes the A
It'll show up in your Q&A section like this:
Next time someone asks the same thing, they get your improved answer.
Why Q&A Pairs Work So Well
The AI understands intent, not just exact words. Train it once with "What fabric is the blue hoodie?" and it'll also handle:
"What's the blue hoodie made of?"
"Material for the blue sweatshirt?"
"Blue hoodie fabric info?"
"Is the blue one cotton?"
Best Q&A Strategies
Start with your FAQ: If people ask it on your website, they'll ask your bot too.
Cover variations: Add Q&As for formal and casual ways of asking the same thing.
Be specific: "The delivery time is 3-5 business days" beats "Pretty fast shipping."
Update regularly: When policies change, update your Q&As immediately.
Think like customers: They ask "How do I return this?" not "What is your return procedure?"
Testing Your Training Data
The AI Agent Window on the right side is your testing playground. Ask questions to verify your bot learned everything correctly.
This testing is completely free - no credits used, so test as much as you want.
What to Test After Adding New Data
After uploading files:
Ask about specific details from those documents
Test both simple and complex questions
Verify the bot can find info buried in long PDFs
After adding text entries:
Ask about the temporary info you added
Test related questions customers might ask
Make sure the bot prioritizes current info over old website content
After creating Q&As:
Ask the exact question you trained
Try different ways of asking the same thing
Test edge cases and similar but different questions
Sample Test Questions
"What are your store hours?"
"How long does shipping take?"
"What's your return policy?"
"Is the blue hoodie available in size large?"
"When do you restock sold-out items?"
Troubleshooting Common Issues
Bot Giving Outdated Information
Check your text entries - delete old temporary info
Update Q&As that reference old policies or pricing
Retrain website pages if they've changed since last crawl
Bot Can't Find Specific Product Details
Upload product catalogs as files if your website lacks detail
Add Q&As for your most commonly asked product questions
Use text entries for specs that change frequently
Answers Sound Too Generic
Add more specific Q&As with exact answers you want
Upload detailed documentation with your preferred language
Use text entries to add personality and brand voice examples
Bot Contradicts Itself
Review all data sources for conflicting information
Prioritize Q&As - they override other sources for specific questions
Update text entries to match your current policies
Managing Your Data Sources
Keep Things Current
Review text entries monthly and delete outdated info
Update Q&As when policies change
Re-upload files when you have new versions
Organize by Priority
Q&As override everything else - use for your most important answers
Text entries are great for current info that might change
Files work best for detailed, stable information
Monitor Performance
Check which questions customers ask most often
Add Q&As for repeat questions the bot struggles with
Use the Improve Answer feature regularly
Remember: The more specific and current your training data, the better your bot performs. Start with the basics, then refine based on real customer interactions.
What Data Sources do
Data Sources are the knowledge your agent can rely on. They do not control behavior or tone. That is handled by your system prompt in Settings. Think of Data Sources as “what the agent knows” and the system prompt as “how the agent should use that knowledge.”
The Four Ways to Feed Your Bot Knowledge
1. Website (Your Main Knowledge Base)
This is where 90% of your bot's smarts come from.
How to add your website:
Enter your homepage URL so Spur can follow internal links across your site. Example: use
sony.comrather thansony.com/product/ps5_controllerClick Crawl to let Spur discover and train on every reachable page under that root.
Use Add link if you only want to add a single page.
Each crawled page shows a status:
Available = Bot knows this content
Processing = Still learning
Failed = Something broke, try recrawling
Pro tip: Use crawl for the whole site, add link for extra pages you want to add.
2. Files (PDFs)
Got product specs? Return policies? Brand guidelines? Upload them here.
What to upload:
Product catalogs
Sizing charts
Shipping info PDFs
FAQ documents
Your brand voice guide
Just drag and drop or click to upload. The bot reads these and uses them to answer questions.
Keep .PDF files under 100MB for smooth processing.
3. Text (Quick Updates)
This is for temporary stuff you dont want to make a whole PDF for.
Perfect for:
"We're closed Dec 24-26 for holidays"
"Shipping delayed 2 days due to weather"
"Flash sale ends midnight 27th Dec"
Store hours and location info
Type it in, hit save. When the info's outdated, just delete or update it. Simple.
4. Q&A (Specific Answers)
Got questions people ask 50 different ways but the answer's always the same? Q&A is your friend.
How to use:
Click Add New Q&A
Type the question (like "What fabric is the blue hoodie?")
Type the answer ("100% organic cotton")
The bot's smart enough to understand when someone asks "whats the blue hoodie made of" or "material for blue sweatshirt" - it'll give them your cotton answer.
Testing Your Data
See that AI Agent Window on the right? That's your playground. Ask it questions to test if it learned your data correctly. This doesn't use credits, so test away.
What to test:
Product questions
Shipping info
Return policy
Store hours
Anything you uploaded
If it's answering wrong, check which data source needs fixing. Maybe recrawl a page or add a Q&A for that specific question.
Quick Troubleshooting
Bot giving outdated info?
Hit Retrain on pages that changed
Delete old text entries
Update your Q&As
Not finding product info?
Make sure you crawled the whole site, not just homepage
Check if product pages are showing "Available" status
Answers too generic?
Add more specific Q&As
Upload detailed product docs
Add text entries with exact info you want shared
Remember: Data sources = what your bot knows. The better the data, the smarter your bot. Feed it well and it'll handle customers like a champ.