Build Knowledge Base Upload, Ingest & Reingest Docs

In this series — Part 2 of 8 (AI Assistant). · Previous: Wapilot AI Assistant Overview: Knowledge Base v… · Next: Knowledge Base Chat (L1): Grounded Answers from… · Series start

Before you start: If you skipped earlier steps, read Wapilot AI Assistant Overview: Knowledge Base v… first so nothing here feels confusing.

Your AI is only as smart as the documents you feed it. To build knowledge base quality, upload truth — not a junk drawer of PDFs.

Teams dump outdated price sheets, conflicting policies, and eighty-page archives, then wonder why WhatsApp answers hallucinate. Upload without ingest leaves files ornamental. Skipping reingest after a price change ships confident wrong numbers.

In this guide, you'll discover what to upload, what to leave out, ingest versus upload, and reingest when content changes — without treating the KB as a shared drive. As a WAPilot team that reviews KB launches weekly, we see current single-topic docs beat giant stale binders.

Soft goal: only ready sources power live answers. Layer map stays L1 truth first; Smart Assistant behavior comes after docs are trustworthy.

Operators who rush this step usually rebuild the same journey twice after customers complain.

Keep the path short enough that a new agent can explain it in one minute during standup.

How do you build knowledge base content that stays accurate?

Sources: Meta's Cloud API get-started guide and template messaging docs. Upload current, clear single-source docs — FAQs, pricing, policies, how-tos, SOPs — not outdated or conflicting files. Ingest after upload so content is searchable; only ready sources should power live answers. When prices or policies change, replace the file and reingest, then spot-check top FAQs.

Leave out outdated PDFs, internal gossip, conflicting price sheets, and huge unrelated archives. Bad docs train confident wrong answers. Prefer one topic per file so humans can maintain and debug.

Layer map: Wapilot AI Assistant overview. Test grounded chat

What is the upload → ingest → reingest loop?

Create or open the knowledge base, upload supported files, name them for humans, then ingest so the assistant can retrieve chunks. Upload ≠ ready. Watch status until sources are ready. On change: replace file, reingest, re-ask the top FAQs that touch that doc.

Attributed example, from a WAPilot SaaS workspace (2026): they changed plan prices but forgot to reingest. The bot quoted old tiers for two days. A reingest checklist after every pricing edit stopped the miss.

1. Upload current single-topic docs only.
2. Ingest until ready before go-live.
3. Exclude conflicting or stale files.
4. Reingest on every material change.
5. Spot-check top FAQs after reingest.

Rule sources: Meta Cloud API get-started and template messaging docs. Wapilot Knowledge Base retrieves your docs; Meta still governs how answers are delivered on WhatsApp.

What belongs in the KB versus elsewhere?

Customer-facing truth belongs in the KB. Internal drafts and gossip do not. Approved scripts can live alongside FAQs, but policies must stay single-source. Org context matters — build inside the right Wapilot organization.

What mistakes create confident wrong answers?

Junk uploads, skip ingest, skip reingest, one giant PDF, and conflicting prices. How we know: WAPilot KB reviews weekly map wrong answers to missing reingest — against Meta’s public Cloud API delivery context, not unofficial “train on the whole internet” claims.

Name files so a new teammate can find the shipping policy in ten seconds.

Quality in, quality out — treat the KB like a product surface, not a dump folder.

After ready status, test in KB chat before WhatsApp go-live.

Smart Assistant playbooks should point at KB truth, not invent parallel prices.

Series 10 next: Knowledge Base Chat habits for grounded answers.

Treat this guide as an operations checklist, not a one-time read. Print the ordered steps and assign an owner for each gate before you flip traffic live.

When something fails in production, capture the customer wording, the node that fired, and the Meta-facing symptom before you redesign the whole canvas.

Keep internal links few and intentional — adjacent series posts and the hub beat random keyword stuffing across twenty URLs.

Respect quiet hours and staffing plans. Automation that earns replies without inbox coverage creates the same unread spikes as careless campaigns.

Version names, changelogs, and a fifteen-minute post-publish review prevent mystery drafts from becoming the source of truth.

If quality rating slides after a launch, pause broad triggers first, fix consent and relevance, then warm traffic again on official Cloud API.

Document product paths clearly for your team: Automation → drip sequences for timed nurture, Flow Builder for interactive journeys, Forms published to Meta for structured intake, AI Knowledge Base versus Smart Assistant for open questions, CRM pipeline beside inbox, and calling under rabbit-calls / WhatsApp calling when voice is needed.

Rehearse the bot-to-agent handoff on a test number: summary present, owner assigned, customer not left in silence. Shared numbers still share one customer care window.

Keep these phrases in your team glossary so search and training stay aligned: upload documents; single source of truth.

FAQs

Is upload enough to go live?

No — ingest until sources are ready, then test answers before WhatsApp.

How often should we reingest?

Whenever prices, policies, or SOPs change — plus a monthly spot-check.

Can one PDF hold everything?

It can, but maintenance and debugging get harder — prefer topic files.

Does a bot-to-agent handoff reset the window on a shared number?

No. One customer thread, one service window. Two brands on one Cloud API line share that clock — split numbers before you scale support and promos on the same line.

Recap: ship the habit, keep a human exit, and measure what customers actually do. When you are ready, start a free trial and grow on official Cloud API.