Knowledge Base

Path: /dashboard/ (Data tab) · Permission: Knowledge Base module

The Knowledge Base is where you train your AI assistant. It defines what the bot knows and can answer using your own documents, webpages, and curated FAQs. Everything you upload is indexed into a vector search engine so the bot can search by meaning — not just keywords. This is the foundation of your bot's intelligence.

How vector search works

Unlike traditional keyword search, FRENZY.BOT uses semantic search (also called vector search). This means:

A user asking "What's the price?" will match a document about "pricing plans" — even though the exact words differ.
The system understands meaning, not just keywords.
Each piece of content is converted into a numerical "embedding" that captures its meaning.
When a user asks a question, the system finds the most semantically similar content and sends it to the AI model for response generation.

This is why the quality of your knowledge base content directly determines the quality of your bot's answers.

Files

Upload internal documents directly to your bot's knowledge base.

Supported formats

Format	Best for
PDF	Policies, reports, product sheets (use text-based PDFs, not scanned images)
DOCX	Internal documents, SOPs, guides
TXT	Plain text content, logs, notes
CSV	Structured data, pricing tables, product lists
JSON	API documentation, configuration data
Markdown	Technical docs, README files
PPTX	Presentations with text content

Step-by-step: Upload a file

Go to Knowledge Base → Files.
Drag files into the upload area or click Upload.
Wait for the status indicator to show Live.
Ask the bot a question from that file to confirm it learned the content.

File management

Enable/disable — Toggle individual files without deleting them.
Preview — View extracted text to verify content quality.
Delete — Remove outdated documents (also removes from the vector index).
Batch operations — Select multiple files for bulk actions.

File size and limits

Limit	Value
Maximum file size	50 MB per file
Recommended file size	Under 10 MB for fastest processing
Number of files	No hard limit — depends on server resources
Processing time	10–60 seconds per file; large PDFs may take several minutes

Best practices

Use clean, text-based PDFs. Scanned image PDFs without selectable text will produce poor results. If you only have scanned PDFs, run them through OCR software first.
Name files descriptively (e.g., pricing_2026.pdf, return_policy_v3.docx) — this helps you identify content later.
Break large documents into smaller files for faster indexing and more precise retrieval. A 200-page manual should be split into chapters.
Remove outdated content to prevent the bot from giving stale answers. Old pricing sheets or expired policies are common culprits.
Review extracted text — Use the Preview feature to verify that the system correctly extracted text from your file.

URLs

Teach your bot directly from public webpages. The system fetches each page, strips navigation, menus, and scripts, and indexes only the readable content.

URL modes

Single page — Add one URL at a time for specific pages.
Sitemap — Provide a sitemap.xml URL to discover and import all pages from a website.

Step-by-step: Add a URL

Go to Knowledge Base → URLs.
Paste a public URL (must be http:// or https://).
Click Add URL.
Monitor the status: Pending → Processing → Live.

Crawl management

Status tracking — Each URL shows its current state (Live / Processing / Error).
Re-crawl — Re-run crawls when your website content changes to keep the bot up to date.
Error handling — If a URL fails, check if the site blocks bots or requires authentication.

Tips for URLs

Public pages only. Pages behind login walls will fail to index.
Sitemap import is the fastest way to index an entire website at once.
Some sites block crawlers. If a URL shows an error, try providing the content as a file instead.
JavaScript-heavy sites (SPAs) may return empty content. If a URL shows no text after indexing, download the page content and upload as a file.
Re-crawl regularly if your website content changes frequently — set a reminder to refresh monthly.

Data Sources

Path: /dashboard/data-sources

The Data Sources page provides an overview of all content sources feeding your knowledge base. It shows:

Column	What it shows
Source	File name, URL, or FAQ identifier
Type	File, URL, or FAQ
Status	Live (searchable), Processing, Error, or Disabled
Size	Content size or number of chunks indexed
Last Updated	When the source was last indexed or re-crawled

WordPress sync

If your website runs on WordPress, FRENZY.BOT can automatically sync content from your WordPress site:

Go to Data Sources or Knowledge Base → URLs.
Add your WordPress site's sitemap.xml URL (usually https://yoursite.com/sitemap.xml).
The system discovers and indexes all published pages and posts.
When you publish new content on WordPress, re-crawl the sitemap to keep the bot updated.

Keep your bot in sync with your website

After publishing new blog posts or updating product pages, click Refresh Knowledge Base to re-index everything. This ensures the bot always has your latest content.

FAQs

Add exact answers for your highest-priority questions. FAQs give you complete control over specific responses.

Why FAQs are powerful

Priority answers — FAQ responses take precedence over document-based retrieval for matching questions.
Instant indexing — FAQs are available immediately, no processing delay.
Precise control — You write the exact answer the bot will use.
Easy updates — Edit or delete individual Q&A pairs without re-uploading files.

Step-by-step: Add an FAQ

Go to Knowledge Base → FAQs.
Click Add FAQ.
Enter the question your users would ask.
Enter your exact answer.
Save — the FAQ is live immediately.
Test in chat to verify.

When to use FAQs

Scenario	Example
Pricing questions	"What does the Pro plan cost?" → Exact pricing answer
Policy responses	"What's your return policy?" → Official policy text
Hours / contact info	"When are you open?" → Business hours
Common objections	"Why should I choose you over X?" → Positioning answer
Corrections	Bot gets something wrong → FAQ overrides with correct answer

Refreshing knowledge

Click Refresh Knowledge Base to re-index all content. This rebuilds the vector index from your current files, URLs, and FAQs.

When to refresh:

After uploading or updating files
After website content has changed
After adding multiple FAQs in batch
After removing outdated content

Refresh is non-destructive

Refreshing does not delete your source content. It re-processes everything to ensure the vector index matches your current data.

Per-bot data isolation

In a multi-bot setup, each bot has its own separate vector collection. This means:

Bot A's knowledge base is completely invisible to Bot B.
Uploading a file to one bot does not affect any other bot.
Deleting a bot removes its entire vector collection and all uploaded files.
Each bot can have completely different content, optimized for its specific audience.

FAQ

Q: A file shows "Error" status.

The file may be corrupt, password-protected, or in an unsupported format. Try converting to PDF or DOCX and re-uploading.

Q: URLs are blocked or show empty content.

Some websites block automated crawlers. Try a different URL, provide a sitemap, or upload the page content as a file instead.

Q: The bot answers incorrectly about a topic.

Add an FAQ with the exact correct answer — FAQs take priority. Also review your system prompt in Settings → Model Behavior for conflicting instructions.

Q: How long does indexing take?

Files: typically 10-60 seconds depending on size. Large PDFs or sitemaps with many pages can take several minutes. Monitor progress in Sync Jobs.

Q: Can I index content behind a login?

Not directly. Pages requiring authentication will fail. Export the content and upload it as a file instead.