Knowledge Bases
Lodgestory CRM › Settings › Knowledge Bases
Give your AI a library. Upload your policies, guides, and rates — connect a live data source — and let a Knowledge Agent answer customers from your own material instead of guessing.
TL;DR
- What it is — a place to store your own documents (policies, FAQs, house guides, rate cards) and, optionally, connect a read-only data source. A Knowledge Agent in a bot journey searches this material and answers from it, with references back to what it read.
- Who it's for — Account Owners and Admins build and curate Knowledge Bases. The Knowledge Agent consumes them on every channel a journey runs on.
- Top outcome — answers grounded in your content — accurate cancellation terms, correct check-in times, real availability — instead of a generic model guessing.
At a glance
| Plan tiers | All paid tiers. AI usage is part of your plan allowance. |
| Who can use it | Account Owners and Admins create and manage Knowledge Bases and data sources. The Knowledge Agent consumes them at runtime. |
| Channels | Channel-agnostic. One Knowledge Base can power a Knowledge Agent on WhatsApp, Email, Instagram, or Web Chat. |
| Integrations | Powers the Knowledge Agent in Bot Journeys. Can connect a read-only external database as a live data source. |
| Supported files | Text, Markdown, HTML, CSV, and question-and-answer lists upload instantly. PDF, Word, PowerPoint, and scanned documents are processed in the background. Up to 50 MB per file. |
| API | Managed inside the CRM. No separate public API. |
How to find it
Sidebar: Settings → Knowledge Bases.
Direct URL: https://app.lodgestory.com/crm/settings/knowledge-bases
[SCREENSHOT: knowledge-bases-landing.png — list of Knowledge Bases with document and data-source counts]
What is Knowledge Bases?
The problem it solves
An AI that answers from its own general training will sound confident and be wrong about your business — it doesn't know your cancellation window, your pet policy, or tonight's rate. You could paste all of that into a system prompt, but prompts have limits, they get unwieldy, and they can't hold a 40-page house manual or a live availability table.
A Knowledge Base fixes this. You upload the material once. When a customer asks a question, the Knowledge Agent searches your material for the passages that matter, reads them, and answers using your wording — and it can pull live facts (a booking status, a room rate) from a connected data source at the same time.
What you get
- Upload what you already have. Drop in policy PDFs, house guides, FAQ lists, rate cards, or a question-and-answer sheet. Text and Markdown are ready instantly; PDFs and scanned documents are read and indexed in the background.
- Grounded answers. The agent answers from the passages it actually retrieved, so replies match your real terms — not a plausible-sounding guess.
- A document inspector. Open any document to see exactly what Lodgestory extracted from it (what the agent reads), the individual passages it was split into for search, and download the original. No more wondering whether your scanned PDF came through cleanly.
- Live data, safely. Connect a read-only database as a data source. Use a reference source for non-personal data (catalogs, rates, availability) the agent can query freely, or a customer source for personal records where every lookup is automatically scoped to the person messaging you — so one guest can never see another's booking.
- Multiple libraries. Keep a "Policies" base, a "Property Guides" base, and a "Rates" base separate, and point each Knowledge Agent at the ones it needs.
How it's different
- Answers cite what they read. The agent retrieves specific passages and answers from them, rather than free-associating. If the material doesn't cover something, a well-instructed agent says so instead of inventing an answer.
- Your files stay usable. The inspector shows the extracted text and the searchable passages, and keeps the original file downloadable — so you can verify quality and fix a bad source.
- Governed live data. Customer-scoped lookups are locked to approved queries and bound to the verified identity of the person in the chat. The agent can't run arbitrary queries against personal records, and it can't look up someone else.
Customer scenarios
- Policy concierge. A villa group uploads its cancellation policy, check-in guide, and Wi-Fi/pool FAQ. A WhatsApp Knowledge Agent now answers "what's your cancellation window?" with the exact terms from the uploaded policy — not a guess.
- Live booking lookups. A hotel connects its bookings database as a customer data source with one approved lookup: "my current booking." When a guest asks "what time is my check-in?", the agent looks up their booking (scoped to their verified phone number) and answers — without exposing anyone else's records.
- Rate questions. A serviced-apartment operator connects a reference source holding this month's rates. The agent answers "how much for the weekend?" from live pricing while quoting the cancellation terms from an uploaded policy in the same reply.
How it fits with the rest of Lodgestory CRM
- Bot Journeys — the AI Agent step, set to Knowledge Agent mode, is what consumes a Knowledge Base. This is the only place a Knowledge Base is used.
- AI Agents — the persona/tools configuration surface for conversational AI. A Knowledge Agent adds retrieval and multi-step reasoning on top of that idea.
Core concepts
| Term | What it means |
|---|---|
| Knowledge Base | A named library of documents (and optional data sources) that a Knowledge Agent can search. You can have several. |
| Document | A file or text you uploaded — a policy, a guide, an FAQ, a Q&A list. |
| Passage | A document is split into smaller passages so the agent can retrieve just the relevant part of a long file. The inspector calls these chunks. |
| Data source | A read-only external database connected to a Knowledge Base, so the agent can answer from live data. |
| Reference source | A data source with no personal data (rates, catalogs, availability). The agent can query it freely. |
| Customer source | A data source with personal records. The agent can only run approved lookups, each automatically scoped to the messaging customer's verified identity. |
| Approved lookup | A pre-defined, admin-written query on a customer source. The agent picks from these; it can't write its own. |
Quick Start — your first Knowledge Base in 10 minutes
Step 1 — Open Knowledge Bases
Sidebar: Settings → Knowledge Bases. Click + New Knowledge Base, give it a name (e.g. Policies) and an optional description. Save.
[SCREENSHOT: knowledge-bases-qs-1-create.png]
Step 2 — Upload a document
Open the Knowledge Base and click Upload document. Pick a file — a policy PDF, an FAQ text file, whatever you have. Text and Markdown appear instantly; a PDF shows a Processing… row while it's read in the background, then becomes a normal document.
[SCREENSHOT: knowledge-bases-qs-2-upload.png]
Step 3 — Check what came through
Click the document to open the inspector. The left pane shows the text Lodgestory extracted (what the agent reads); the right pane shows the individual passages it was split into. If a scanned PDF came out garbled, you'll see it here — re-upload a cleaner copy.
Step 4 — (Optional) Connect a data source
Open the Data sources tab and click Add source. Choose Reference (non-personal — the agent can query freely) or Customer (personal records — you'll add approved lookups, each scoped to the messaging customer). Enter read-only database details and save.
Step 5 — Point a Knowledge Agent at it
Open Bot Journeys. Add an AI Agent step, switch it to Knowledge Agent, and tick this Knowledge Base. Write the persona/instructions, save, and deploy.
Step 6 — Test
In the journey's test panel, ask a question your document answers. The agent should search the Knowledge Base, read the relevant passage, and reply using your wording.
What's next
- Add more documents; keep policies and guides in separate bases if they serve different journeys.
- Connect a live data source for booking lookups.
- Re-index after a bulk update (see below).
How it works
When you upload a document, Lodgestory reads it (including running scanned PDFs and images through text recognition), breaks it into passages, and indexes those passages for meaning-based search. Text and Markdown are ready in a second; large or scanned files process in the background and show a Processing… row until they're done — reload-safe, so you can navigate away and come back.
At runtime, when a customer asks a question, the Knowledge Agent searches your indexed passages for the most relevant ones, reads them, and — if a data source is connected — can also fetch live facts. It can take several steps (search, read, look something up, then answer), and it replies using the material it retrieved. If nothing in your material covers the question, a well-instructed agent says so rather than inventing an answer.
Customer-scoped data lookups are the careful part: the agent can only run the approved lookups you defined, and each one is automatically filled with the verified identity of the person in the chat. The agent never writes its own query against personal records and can't look up a different customer.
Features in depth
Knowledge Base list
Each Knowledge Base shows its name, description, document count, and connected data-source count. Create, open, or delete from here.
Documents
Upload text, Markdown, HTML, CSV, question-and-answer lists, PDFs, Word, PowerPoint, and scanned files (up to 50 MB each). Text-like files are indexed instantly; binary/scanned files process in the background with a live Processing… (or Failed) row that survives a page reload. Remove a document at any time; the agent stops using it immediately.
Document inspector
Open any document to see:
- Parsed text — the text Lodgestory extracted (what the agent actually reads). The single best way to confirm a scanned PDF came through cleanly.
- Passages — the individual pieces the document was split into for search, with the count and the indexing status.
- Download original — the file exactly as you uploaded it.
- Delete — remove the document from the Knowledge Base.
[SCREENSHOT: knowledge-bases-document-inspector.png — split view: parsed text on the left, passages on the right]
Data sources
Connect a read-only external database:
- Reference — non-personal data (rates, catalogs, availability). The agent can run its own read-only queries against it.
- Customer — personal records. You define one or more approved lookups (e.g. "my bookings"); the agent can only run those, and each is automatically scoped to the messaging customer's verified identity.
For safety, customer-scoped lookups are withheld on channels where the sender's identity can't be verified (for example, inbound email), while reference lookups stay available.
Re-index
After a large update, re-index a Knowledge Base to rebuild its search index. It runs in the background with no downtime — the agent keeps answering from the current index until the new one is ready.
Roles and permissions
| Action | Account Owner | Admin | User | Bot |
|---|---|---|---|---|
| View Knowledge Bases and documents | Yes | Yes | No | No |
| Create or delete a Knowledge Base | Yes | Yes | No | No |
| Upload or delete a document | Yes | Yes | No | No |
| Connect or remove a data source | Yes | Yes | No | No |
| Use a Knowledge Base at runtime | — | — | — | Yes (via a Knowledge Agent step) |
Limits a user will run into
| Limit | Value |
|---|---|
| File size per upload | 50 MB |
| Instant (inline) text upload | Around 2 MB; larger text files process in the background |
| Supported instant formats | Text, Markdown, HTML, CSV, JSON, Q&A lists |
| Supported background formats | PDF, Word, PowerPoint, images / scanned documents |
| Data source access | Read-only. Customer sources run approved lookups only. |
Errors and FAQ
Errors you might see
| Message | What it means | What to do |
|---|---|---|
| Binary documents need a storage bucket | A PDF/Word upload needs a storage location configured. | Text/Markdown upload without one; contact support to enable file storage. |
| The uploaded file is empty | The file had no readable text. | Check the file opens correctly and re-upload. |
| A document sits on Failed | The file couldn't be read (e.g. a corrupt or password-protected PDF). | Dismiss the row and re-upload a clean, unlocked copy. |
| A document stays on Processing… for a long time | A large scanned document is being read; text recognition on many pages takes time. | Leave it — it's reload-safe. If it never completes, re-upload. |
| Customer-tier data sources need at least one approved lookup | You saved a customer source with no lookups defined. | Add at least one approved lookup before saving. |
FAQ
What file types can I upload?
Text, Markdown, HTML, CSV, JSON, and question-and-answer lists index instantly. PDFs, Word, PowerPoint, and scanned documents are read and indexed in the background. Up to 50 MB each.
Why does my PDF say "Processing…"?
Scanned or image-based PDFs are run through text recognition page by page, which takes a little time. The row is reload-safe — you can leave and come back.
How do I know the agent read my document correctly?
Open the document inspector. The Parsed text pane shows exactly what Lodgestory extracted — if a scan came through garbled, you'll see it and can re-upload a cleaner copy.
Can the agent see another customer's records?
No. Customer data sources only run the approved lookups you define, and every lookup is automatically scoped to the verified identity of the person in the chat. The agent can't write its own query or look up anyone else — and these lookups are withheld entirely on channels where identity can't be verified.
Do I need a Knowledge Base to use AI at all?
No. A plain Conversational AI Agent works without one. Add a Knowledge Base when you want answers grounded in your own documents or live data.
I updated a lot of documents — do I need to re-index?
Individual uploads and deletes update the index automatically. Re-index is only for a full rebuild after a large change; it runs with no downtime.
Changelog
- Jul 2026 — Initial release: Knowledge Bases, the document inspector (parsed text + passages + download), external reference/customer data sources, and the Knowledge Agent step in Bot Journeys.
Related modules and next steps
- Bot Journeys — add an AI Agent step in Knowledge Agent mode to use a Knowledge Base.
- AI Agents — personas and tools for conversational AI; the Knowledge Agent builds on the same ideas with retrieval and multi-step reasoning.