RAG stands for "Retrieval Augmented Generation". In simple terms, it's how your agent learns information specific to your business.
🎯 Imagine this:
You run a shoe store and want your agent to know about all your products. Instead of having the agent guess or give generic information, you upload your full catalog. Now, when a customer asks "Do you have waterproof boots?", the agent will search that catalog and answer accurately.
✨ Result: Precise answers based on YOUR information, not generic internet knowledge.
Which documents can you upload?
Product catalogs, user manuals, company policies, info brochures.
TXT
FAQs, sales scripts, price lists, product information.
MD (Markdown)
Technical documentation, how-to guides, blog posts, knowledge base articles.
DOCX (Word)
Word documents with company information, procedures, reports.
How to upload documents:
Go to the agent configuration
Click "Configure" on your agent's card.
Select the "Knowledge" tab
It's the second tab in the top menu.
Drag your files or click "Upload"
You can upload multiple files at once.
Wait for them to be processed
The system reads and organizes the information automatically.
Done! Test by asking about that information
Go to the chat and ask questions related to the documents.
⚠️ Important:
- • Documents must be in a language the model can read (Spanish, English, etc.).
- • Very large files may take longer to process.
- • The information must be readable text (PDFs scanned as images won't work well).
If you have information on your website that changes frequently (such as prices, availability or new products), you can configure URLs for the agent to read it automatically.
💡 Ideal use cases:
- • Company blog: your agent will always know your latest articles.
- • Online catalog: it will always have up-to-date products and prices.
- • FAQ page: if you update your FAQ on the site, the agent will know.
- • Technical docs: keep the agent updated with the latest version.
How to configure it:
- 1. In the Knowledge tab, find the "Web Sync" section.
- 2. Paste the URL of the page you want the agent to read.
- 3. Configure how often it should update (daily, weekly, etc.).
- 4. Save your changes.
Practical example:
You have a products page on your website. The agent will read it and learn about all listed products. When you update that page, the agent will automatically know.
If you upload many documents (more than 10–20 files), it's recommended to enable the Vector Database. This helps the agent find relevant information much faster and more accurately.
🔍 How does it work?
Instead of looking for exact words, the system understands the meaning of questions and finds the most relevant information, even if it doesn't use the same words.
❌ Without Vector Database:
Question: "How much does a sofa cost?". If your docs say "couch price" instead of "sofa", it might not find it.
✅ With Vector Database:
Question: "How much does a sofa cost?". It understands that sofa = couch = living room furniture, and finds the right information.
How to enable it:
- 1. In the Knowledge tab, under Search Configuration, enable the "RAG (Retrieval Augmented Generation)" switch.
- 2. Enable the "Semantic Search" and/or "Keyword Search" options depending on what you need.
- 3. Optional: you can also enable "Hybrid Search" for maximum accuracy.
- 4. Save your changes.
💡 Hybrid Search
Combines smart search with traditional keyword search. It's the most powerful option but uses a bit more resources. Recommended if you have very technical or specific information.
- ✓
Organize information before uploading it
A well-structured document with clear headings and sections works much better than a messy file.
- ✓
Less is more (but complete)
It's better to have 5 well-crafted documents than 50 with repeated or contradictory information.
- ✓
Use clear, direct language
Avoid overly technical jargon unless it's necessary. Write as if you were explaining it to a real customer.
- ✓
Update regularly
If your prices change or you add new products, update the documents so the agent has correct information.
- ✓
Test after uploading information
Whenever you add documents, ask the agent about that topic to verify it understood correctly.