Back to DocumentationSwanDigitals Docs

AI Training & Knowledge Base

Train your SwanAI chatbot to answer customer queries accurately by building a knowledge base. The AI uses RAG (Retrieval Augmented Generation) to provide context-aware responses from your documents.

AI Providers

SwanDigitals supports multiple AI providers. Switch between them anytime based on your needs.

Primary AI Engine

Fast, cost-effective, tuned for Hindi and Indian context.

default

Alternative AI Engine

Robust cloud-based alternative with excellent general-purpose reasoning.

alternative-cloud-model

Advanced Reasoning Engine

Best for nuanced, detailed responses. Available on higher plans.

advanced

Switching AI: Go to Settings → AI → Provider to change your AI model. Configuration changes apply immediately to all conversations.

Supported Document Types

📄

PDF

Product manuals, policies

📝

Word (.docx)

FAQs, guides

🔗

URLs

Website pages (auto-scraped)

📋

Text (.txt)

Raw content

Building Your Knowledge Base

Method 1: Manual Upload

  1. Navigate to AI → Knowledge Base
  2. Click Add Documents
  3. Select files from your computer or paste URLs
  4. Add tags for organization (e.g., "pricing", "support", "products")
  5. Click Upload & Index
  6. Wait 1-5 minutes for indexing to complete

Method 2: Website Crawler

Automatically crawl your website and import all content into the knowledge base.

  1. Go to AI → Knowledge Base → Import from URL
  2. Enter your website URL (e.g., https://swandigitals.com)
  3. Set crawl depth (how many clicks deep to go)
  4. Exclude paths: /blog, /pricing-old
  5. Click Start Crawl
  6. Review imported content and remove irrelevant pages

Method 3: Programmatic Build

Use the KB Builder API for automated knowledge base updates.

# Python example with SwanAI SDK
from swanai import KBBuilder

builder = KBBuilder(api_key="your_key")
builder.add_urls([
    "https://example.com/faq",
    "https://example.com/pricing"
])
builder.add_files(["./policy.pdf", "./faq.docx"])
builder.build()

Multilingual Training

Train the AI to respond in Hindi, regional Indian languages, and Hinglish.

Training for Hindi Responses

  1. Add documents with Hindi content or Hindi-English bilingual content
  2. Navigate to AI → Settings → Language
  3. Select Hindi as primary response language
  4. Enable Hinglish mode for mixed Hindi-English responses
  5. Test with queries like "मूल्य क्या है?" or " kitna price hai?"

Supported Languages

English
Hindi
Marathi
Tamil
Telugu
Kannada
Gujarati
Bengali
Malayalam
Punjabi

Testing & Optimization

AI Playground

Test how your chatbot responds to different queries before going live.

  1. Go to AI → Playground
  2. Type a customer question
  3. Review the AI response and source citations
  4. If answers are poor, add more relevant documents or improve existing content

Tip: Use specific terms from your industry. The AI learns from your documents, so include domain-specific terminology for better accuracy.

Next Steps

  • Voice AI — Enable AI-powered phone support
  • Automation — Set up routing rules and escalation logic
  • Platform Guide — Learn more platform features