Knowledge base & document ingestion.

The KB is what the bot uses to answer mid-conversation questions. Different from qualifying questions: KB answers questions visitors ask you (about pricing, process, objections), while qualifying questions are what the bot asks them.

What goes in the KB

Pricing — your real pricing, not “contact us”
Process — what happens after they sign up / book / pay
Objection-handling — common reasons people hesitate, with your honest answer
Product details — what’s included, what’s not
Differentiation — what makes you different from alternative X
Proof points — case studies, customer counts, results
Process objections — “do I need to sign a contract?”, “can I cancel anytime?”
Edge cases — unusual situations you’ve handled before

Adding a manual entry

Knowledge Base tab → Add a manual entry:

Field	Notes
Title	Short label (e.g. “Refund Policy”, “Onboarding Process”)
Content	The actual text the bot will use as context
Category	Pick from: product, company, pricing, process, objection, other

The category is for your own organisation — the bot doesn’t filter by category at retrieval time.

Edit — inline form on each row. Changes are searchable immediately (no reindex wait). Delete — removed from the search index immediately.

Bulk import (CSV / XLS / XLSX)

Knowledge Base tab → Bulk Import. Upload a spreadsheet with columns:

title	content	category
Pricing	We charge $99/mo per user, billed annually. No setup fees.	pricing
Cancellation	You can cancel anytime — no penalties. Pro-rated refund.	process

Constraints: Max 200 rows per upload. Headers required: title, content (case-insensitive); category optional.

Two silent behaviours to know about

Rows missing either title or content are skipped without raising an error — the response message reports how many rows were imported vs. skipped, so check it if your imported count is lower than your file’s row count.

Any category value that isn’t one of the six valid options (product, company, pricing, process, objection, other) is normalised to empty — the row imports but lands uncategorised. Capitalisation is forgiving (“Pricing” matches), but typos and out-of-enum values like “marketing” are silently dropped from the category column.

Writing KB entries that work

The bot does well when:

Entries are specific and concrete — “$99/month per user, billed annually” not “competitive pricing”
Entries match how visitors actually phrase questions
Tone matches your overall brand voice
One topic per entry — don’t bundle pricing + cancellation + onboarding into one

The bot does poorly when:

Entries are vague marketing copy
Two entries contradict each other
The KB is empty (the bot has nothing to ground on for objections)

Common starter set (10 entries)

For most SMBs, this 10-entry skeleton handles 70%+ of in-chat questions:

Pricing
What’s included
What’s NOT included (gotchas, caveats)
Cancellation policy
Refund policy
Onboarding / time-to-value
Customer support
Common objection #1 (often: “are you cheaper than X?”)
Common objection #2 (often: “how long does it take?”)
Trust signal (case study, customer count, certifications)

Document ingestion (power user)

If you have existing sales content (a help-centre URL, a product PDF, an FAQ doc), you can ingest it directly instead of re-typing into KB entries.

Supported formats

Format	Extension	Notes
URL (single page)	—	Public web page; we extract main content
URL (whole site)	—	Point at a site and we’ll follow its sitemap to ingest multiple pages in one go
PDF	`.pdf`	Text-based PDFs work directly. Scanned/image PDFs are now read via OCR — no need to convert them first. OCR is capped at the first 50 pages per PDF; anything beyond page 50 is silently dropped.
Word	`.docx`	Modern Word format. Legacy `.doc` is rejected — convert to `.docx` first.
Markdown	`.md`	Plain Markdown

How it works

Knowledge Base tab → Add a source. For a URL, paste it and click Import URL. For a file, click ↑ Upload PDF / Word / Markdown and pick the file from your computer. There is no single “Ingest” button — the URL form auto-submits via Import URL, and the file picker auto-submits on selection.

Status flow shown in the Imported documents list:


pending → Processing… → Ready (success) or failed (with an error message)

Scanned PDFs show Processing (OCR)… mid-flight to flag the longer wait.

Behind the scenes Tenlo fetches, parses, chunks into ~500-token blocks, embeds each chunk, and adds to your search index. Typical end-to-end time: 30 seconds to 2 minutes per document.

Sitemap crawl (multi-page URL ingest)

You no longer have to ingest help-centre pages one at a time. Paste a site URL and Tenlo will look for its sitemap and ingest the pages it lists. A crawl counts as one ingest against your hourly limit, no matter how many pages it pulls in.

Limits

100 documents per business at any one time — deleting a document frees a slot. This cap is shared with the Support Chatbot (P01).
25 MB max per file
20 ingests per hour (a sitemap crawl = 1 ingest)

Deleting

The Imported documents list has a Delete button per row. Clicking it removes all chunks from the search index, deletes the file from storage, and removes the document from the list. The bot stops citing it immediately.

KB entries vs. document ingestion

Use KB entries when…	Use document ingestion when…
The answer is short and specific	You have long-form policy or guide content
You want tight wording control	The content already exists and is accurate
The topic is a single Q&A	The doc covers many topics
You’re starting from scratch	You already maintain a help centre or product doc

You can mix both freely. The bot searches everything together.

What’s not yet supported

JavaScript-rendered single-page apps — save the page as a PDF and upload that instead
Scheduled re-ingestion — today, you re-ingest manually after content changes
Multi-file batch upload — ingest documents one at a time

These are roadmap items. (Sitemap crawl and scanned-PDF OCR — previously listed here — are now live.)