Knowledge Base
Ground your AI agents on your proprietary business documents to ensure accurate, hallucination-free answers.
Overview
The Knowledge Base is the central repository of information your AI agents use to answer customer questions. Rather than relying on public internet data (which can be inaccurate or irrelevant), KantaSwara uses a technique called Retrieval-Augmented Generation (RAG). When a user asks a question, the agent searches your uploaded documents for the answer and synthesizes a response.
Supported Formats
- Documents: PDF, DOCX, TXT. Ensure text is selectable (not just scanned images).
- Websites: Provide a URL, and KantaSwara will crawl the text content of that specific page.
- Raw Text: Copy and paste FAQs or snippets directly into the dashboard.
Uploading Documents
- Go to the Knowledge Base tab in your Organization Dashboard.
- Click Add Document.
- Select your file or input your URL.
- Assign a descriptive title and tags (optional).
- Click Upload.
Knowledge Processing (Embeddings)
Once uploaded, your document enters the Processing state. KantaSwara breaks the text down into small "chunks" and converts them into mathematical vectors (embeddings). This allows the AI to perform semantic searches at lightning speed during live phone calls. Processing typically takes less than 60 seconds. Once the status changes to Active, the knowledge is immediately available to your live agents.
Best Practices
- Use Q&A format: Explicit FAQs (Question: X, Answer: Y) are easier for the AI to retrieve perfectly.
- Remove boilerplate: Strip out lengthy legal disclaimers or repetitive headers from your PDFs to improve search relevance.
- Keep it updated: If pricing changes, delete the old document and upload a new one immediately. Agents will instantly use the new data.