Knowledge Base & Custom Pipelines

Turn documents into answers

Vectrant's RAG (Retrieval Augmented Generation) knowledge base ingests your business documents and makes them searchable through natural conversation. Customers and employees ask questions in plain language and get accurate answers sourced from your actual documentation — not hallucinated responses.

Multi-format document ingestion

Upload PDF, DOCX, XLSX, PPTX, Markdown, CSV, and plain text files. Documents are automatically chunked into semantic sections, embedded into vector representations, and indexed for retrieval. New documents can be added through drag-and-drop in the admin dashboard or via automated sync from a watched directory.

Custom pipelines

Build dedicated knowledge pipelines for specific topics — warranty policies, financing options, vendor specifications, training materials, return procedures, delivery schedules. Each pipeline has its own document collection, system prompt, and keyword triggers. Three retrieval types are supported:

• Curated — static content returned directly, ideal for policies and procedures that don't change often
• Documents — uploaded files indexed into a per-pipeline vector collection for semantic search
• Web scraping — URLs fetched and indexed automatically, keeping content in sync with your website

Customer vs. internal isolation

Not all knowledge is meant for customers. Vectrant separates customer-facing knowledge from internal employee documents. Customer queries are routed to a customer knowledge pipeline, while employee and agent queries can access the full internal knowledge base. Role-based access control ensures the right people see the right information.

ERP and purchase order integration

Connect your ERP system to surface real inventory and purchase order data. When a product is out of stock, the AI can tell customers when the next shipment arrives. Employees can query order status, delivery timelines, and vendor information through the same chat interface.

Live website search

When document-based retrieval doesn't have the answer, Vectrant can fall back to live website scraping — searching your public website for relevant pages and extracting the answer in real time. This provides a safety net that ensures customers always get the most current information.