​
What Are Knowledge Bases?

Think of a Knowledge Base as your AI’s reference library. When processing data, AI tasks can consult these documents to provide more accurate, contextually relevant answers with proper citations. How they work:
  • Upload relevant documents (PDFs, manuals, policies)
  • Claro automatically chunks and indexes the content
  • During task execution, AI retrieves the most relevant passages
  • Results include citations linking back to source documents

​
Storage Limits by Plan

PlanKnowledge BasesStorage per KBMax File Size
Free110 MB10 MB
Starter550 MB20 MB
DedicatedUnlimited500 MB*100 MB
*Higher quotas available on request

​
Setting Up Knowledge Bases

Step 1: Create or Select
  • Each project starts with one empty Knowledge Base
  • Create additional KBs from the Knowledge Bases panel
  • Use descriptive names like “Product-Manuals-2024” or “Brand-Guidelines”
Step 2: Add Content
  • Upload PDFs: Drag and drop individual files or entire folders
  • Paste text: Add policies, guidelines, or reference content directly
  • Organize by topic: Keep KBs focused for better retrieval accuracy
Step 3: Connect to Tasks
  • When creating AI tasks, select relevant Knowledge Bases
  • Multiple KBs can be used simultaneously
  • AI automatically finds the most relevant passages

​
Best Practices

✅ Effective Knowledge Base Strategy:
  • Stay focused: One KB per topic (products, policies, brand voice)
  • Keep current: Update documents regularly, archive old versions
  • Use clear naming: Include dates and purpose in KB names
  • Test relevance: Smaller, focused collections perform better than large dumps
✅ Ideal Use Cases:
  • Product specifications for accurate attribute extraction
  • Brand guidelines for consistent content generation
  • Policy documents for compliance and regulatory tasks
  • Technical manuals for troubleshooting and support
⚠️ What to Avoid:
  • Time-sensitive data: Stock prices, inventory levels (use live data sources)
  • Overly broad collections: Mix of unrelated topics reduces accuracy
  • Outdated information: Archive old versions to prevent confusion
💡 Pro Tips:
  • For multilingual dataset, create separate KBs for each language
  • Name KBs with date suffixes for easy version tracking
  • Deleting a KB doesn’t affect existing task results—they simply stop using that context