✅ What is the Knowledge Base?
It’s a collection of content (text, documents, URLs) that the Voice Agent can use to generate accurate and relevant answers, without changing the prompt.
Typical use cases:
Customer support
FAQ and documentation
Technical help
Educational materials and training content
🔍 How It Works
During a call, the agent automatically searches the Knowledge Base.
It scans the call transcript and retrieves the most relevant fragments.
These fragments (“chunks”) are passed as context to the AI model (LLM), which uses them to answer the caller accurately.
📁 Supported Source Types
You can upload:
🔗 Web pages (URLs) – sync website content
📄 Files (.pdf, .docx, .pptx, .txt, .xml, .xls, etc.)
✍️ Manual text – custom pasted content (e.g., definitions, scripts, instructions)
✅ Tip: Use structured and readable formats like Markdown where possible.
Supported file formats:
.bmp, .csv, .doc(x), .eml, .epub, .heic, .html, .jpeg, .md, .msg, .odt, .org, .p7s, .pdf, .ppt(x), .rst, .rtf, .tiff, .txt, .tsv, .xls(x), .xml
⚠️ Limits
Type | Limit |
URLs | max 500 |
Files | max 25 (each ≤ 50 MB) |
Manual text entries | max 50 |
✅ Best Practices
Use clear paragraph structure and headers (Markdown works best)
Group related info into unified documents
Avoid vague terms like “this”, “that”, “the same” without clear context
Include specific names, numbers, models, and measurements
If a page contains too much unrelated information, copy only the needed text
💡 Example: How It Looks in the Panel
In the "AI Voice Agent" creation wizard, go to step 3 – Knowledge Base
There, you’ll see three options:
Add Web Pages
Upload Files
Add Text Manually
