Knowledge Base

Upload your documents, FAQs, product guides, and policies. Your AI reads them and uses the knowledge to answer customer questions accurately — no hallucinations, no generic responses.

How It Works

1. Upload documents — PDFs, text files, or paste content directly
2. AI indexes automatically — extracts text and makes it searchable
3. Conversations get context — when a customer asks a question, the AI checks your knowledge base first
4. Accurate answers — replies cite your content, not generic AI knowledge

Two Types of Knowledge

Documents

Upload files that contain information your AI should know:

• Product catalogs and pricing
• FAQs and help articles
• Shipping and return policies
• Company information and history
• Technical documentation

Supported formats: PDF, TXT, DOCX, and plain text paste.

The AI extracts text from documents and includes relevant sections when answering customer questions.

Instructions

Custom rules that shape how your AI behaves:

• “Always greet customers by name if available”
• “Never discuss competitor products”
• “If asked about refunds, direct to [email protected]”
• “Respond in formal English for business inquiries”
• “Include our website link when discussing pricing”

Instructions are shorter, directive rules. Documents are longer reference material.

Live Integration Data

Your AI doesn’t just read static documents — it pulls real-time data from connected integrations.

When a customer asks “Where’s my order?” or “Do you have this in blue?”, the AI already knows. No manual uploads, no stale data.

Integration	What the AI sees	Example question it handles
Shopify	Products, variants, prices, inventory, customer orders, tracking	”Is the black hoodie in stock in XL?”
HubSpot	Contacts, deals, company info, ticket history	”What’s the status of my support ticket?”
WooCommerce	Products, orders, shipping status	”When will my order arrive?”

This data is injected into the AI context alongside your documents and instructions — the AI uses all of it together to give accurate, personalized answers.

Why this matters:

• Customers get real answers about their orders and your products — not generic responses
• Support volume drops because the AI resolves product and order questions without a human
• No manual data entry or CSV imports — connect once, data stays current

E-commerce businesses using AI with live product data see a 25% cart recovery rate and 60% reduction in “where’s my order” tickets compared to rule-based bots.

Scope & Cascading

Knowledge applies at three levels, cascading down:

Level	Applies to	Use for
Organization	All channels and businesses	Company-wide FAQs, brand guidelines
Business	All channels in that business	Business-specific products, policies
Channel	That channel only	Channel-specific greetings, promotions

A channel inherits knowledge from its business and organization. Channel-level knowledge takes priority when there’s a conflict.

Adding Knowledge

Add Your Website

Paste any URL and the AI extracts every page on your site automatically. Each page becomes a knowledge base document — your product pages, about page, FAQ, pricing, blog posts — all indexed and searchable by the AI in seconds.

1. Go to Knowledge Base → click Add URL
2. Paste your website URL (e.g. yourstore.com)
3. SendHub crawls the page and discovers linked pages on your domain
4. All pages are extracted and saved as documents — AI starts using them immediately

No copy-pasting, no file uploads. Your entire website becomes your AI’s knowledge base.

Connect Your Tools

When you connect an integration, its data becomes part of your AI’s knowledge automatically — no manual setup.

Integration	What gets added	Updates
Shopify	Products, variants, prices, inventory, customer orders, tracking	Real-time — AI always sees current data
HubSpot	Contacts, deals, company info, tickets	Synced on every conversation
WooCommerce	Products, orders, shipping status	Real-time

Your AI combines this live data with your documents and instructions. A customer asking about a product gets the current price from Shopify and your return policy from your FAQ — in one answer.

Upload Files & Write Articles

1. Go to Knowledge Base in the sidebar
2. Click Write Article to create content directly, or Upload File for existing documents
3. Set the scope (org, business, or channel)
4. Save — AI starts using it immediately

Supported formats: PDF, DOCX, TXT, Markdown, CSV (max 10MB)

Via API

// Add a document
await sendhub.from('ai_documents').insert({
  org_id: 'your-org-id',
  title: 'Shipping Policy',
  raw_text: 'Free shipping on orders over $50...',
  status: 'ready',
})

// Add an instruction
await sendhub.from('ai_instructions').insert({
  org_id: 'your-org-id',
  content: 'Always mention our 30-day return policy',
  scope: 'all',
  sort_order: 1,
})

How the AI Uses Knowledge

When a customer sends a message, the AI assembles context from every available source:

Customer: "Can I return the blue hoodie I ordered last week?"
        |
        v
AI assembles context:
  1. Instructions — "Always mention our 30-day return policy"
  2. Documents — Returns policy PDF, FAQ article
  3. Shopify — Order #1847: Blue Hoodie, delivered March 21
  4. Conversation history — previous messages
        |
        v
AI replies: "Hi! Your Blue Hoodie from order #1847
            was delivered on March 21. You're within
            our 30-day return window. I can start
            a return for you — return shipping is free."

The AI combines every connected source in a single reply. A customer asking about a return gets your policy (document), their specific order (Shopify), their CRM notes (HubSpot), and your instructions — all in one answer.

The AI searches your knowledge base semantically — it understands what the customer is asking and pulls the most relevant documents, not just keyword matches. A character budget (50,000 chars) keeps responses fast and focused, even with large knowledge bases.

Best Practices

• Keep documents focused — one topic per document works better than one massive file
• Use clear titles — the AI uses titles to decide which documents are relevant
• Update regularly — change a document and the AI picks it up on the next message
• Test with real questions — send test messages to verify the AI uses your knowledge correctly
• Combine instructions + documents — instructions for behavior rules, documents for reference material

Troubleshooting

AI doesn’t use my knowledge

• Check that the document status is “ready” (processing can take a moment)
• Verify the scope matches the channel you’re testing on
• Make sure the document contains the information the question is about

AI gives wrong answers

• Check for conflicting information across documents
• Add a specific instruction to override: “When asked about X, always say Y”

Document won’t upload

• Maximum file size: 10MB
• Supported formats: PDF, TXT, DOCX
• If a PDF has scanned images (not text), the AI can’t read it