AI Chatbot & RAG Knowledge Base
Train Omni Chat Stack’s multilingual AI chatbot on your website, PDFs, Word docs, and Q&A pairs—answering in 7 languages with grounded RAG retrieval.
7 supported languages (including Roman Urdu)
The AI Chatbot automatically detects the customer’s language and responds naturally in the same language—including Roman Urdu typed in Latin script on WhatsApp. Configure active languages inside Knowledge → Languages Config (/knowledge) to optimize response latency.
| Language | Script / Style | Text AI Chatbot | AI Voice Agent |
|---|---|---|---|
| English | Standard | Yes | Yes |
| Urdu | Nastaliq / Urdu Script | Yes | Yes |
| Roman Urdu | Latin script (e.g., "order kab milega?") | Yes | Yes |
| Arabic | Arabic Script | Yes | Yes |
| Spanish | Español | Yes | Yes |
| Portuguese | Português | Yes | Yes |
| Pashto | Pashto Script | Yes | Yes |
Three ways to train your AI
1. Website Crawler (WebsiteTab)
Enter your website URL and control crawl depth and path inclusions (POST /api/v1/knowledge/scrape). Omni Chat Stack indexes your live pages and auto-generates an AI business persona tailored to your brand.
2. Document Upload (DocumentsTab)
Upload up to 10 files per batch (max 10 MB each) in PDF, DOCX, TXT, MD, or HTML formats (POST /api/v1/knowledge/documents). Documents are split into KnowledgeChunk embeddings stored in Qdrant and MongoDB.
3. Direct Q&A Pairs & Knowledge Entries (QaTab / KnowledgeTab)
Add authoritative fee schedules, return policies, timings, and high-priority FAQ pairs in /knowledge or /settings/ai. Top matching entries are injected directly into the AI system prompt.
Chat Test, Chunk Debugger & Usage Bar
- Chat Test Tab: Simulate customer questions inside /knowledge before going live and verify tone, language detection, and persona.
- Retrieval Debugger: Inspect exact retrieved chunks and similarity scores via POST /api/v1/knowledge/test or GET /api/v1/knowledge/documents/:id/chunks-debug, and re-index updated documents anytime.
- Quota monitoring: The Knowledge Usage Bar tracks indexed documents, chunk counts, and monthly AI conversation usage (300 conversations/mo included on AI Chat and higher tiers).