Pet policy from approved document
Scenario: Visitor asks, “Do you allow small dogs?”
System behavior: AI retrieves the active pet policy document for that community.
Result: AI drafts or sends a property-accurate response based on configuration.
Document Center / AI Knowledge Base
Single source of truth for property knowledge
The Document Center gives each PMC/community a controlled place to manage the documents and approved URLs that staff and AI rely on: pricing, policies, brochures, floor plans, care/service information, move-in requirements, FAQs, amenities, and promotions.
Develop Document Center UI and workflow planDefine versioning/update planS3 + vector searchPMC/community isolation
1. Feature Purpose
Sales directors often maintain the same information across multiple places: websites, brochures, APFM/Caring profiles, internal folders, email attachments, and spreadsheets. This creates inconsistent information and weak AI accuracy.
The Document Center solves this by becoming Journey’s approved property knowledge layer. It is not just file storage; it is the operational knowledge base that powers response drafting, AI chat, internal lookup, onboarding readiness, custom reports, and third-party content consistency.
2. Supported Knowledge Sources
Uploaded documents
- PDF pricing sheets
- Pet policy documents
- Floor plans and brochures
- Move-in requirements
- Care/service documents
- FAQs and promotions
Approved URLs
- Property website pages
- Online pricing/policy pages
- Public amenity pages
- Approved community FAQ pages
- Re-scrape / refresh option
- Last fetched timestamp
URL ingestion should be controlled. Customers provide approved URLs; Journey fetches, extracts, sanitizes, and indexes the content only after validation.
3. Document + URL Workflow
flowchart TD
A[Admin opens Document Center] --> B{Source Type}
B -->|File Upload| C[Upload file]
B -->|Approved URL| D[Add URL]
C --> E[Select category + PMC/community/property]
D --> E
E --> F[Store file or URL metadata]
F --> G[Extract text / scrape page]
G --> H[Sanitize + redact unsafe content]
H --> I[Chunk + create embeddings]
I --> J[Vector search index]
J --> K[Mark AI Indexed]
K --> L[AI retrieval for drafts, chat, reports, lookup]
4. Document Center UI and Workflow Plan
Required planning item: Develop Document Center UI and workflow plan.
Document/URL list
- Name/title
- Source type: file or URL
- Category
- PMC/community/property
- Active version
- AI indexed status
- Last updated / last fetched
- Owner / uploaded by
Management actions
- Upload
- Add URL
- Replace
- Re-scrape URL
- Archive/delete
- Re-index
- Approve for AI
- View version history
5. Versioning / Update Plan
Required planning item: Define Document Center versioning/update plan.
| Item | Plan |
|---|---|
| File replacement | Creates a new version. Previous versions become archived unless restored. |
| URL refresh | Stores last fetched content version and timestamp. |
| Active version | Only active approved version is used by AI retrieval. |
| Last indexed date | Shows whether AI search is current. |
| Freshness alerts | Pricing/policy docs can be flagged for periodic review. |
| Rollback | Authorized user can restore a previous version if needed. |
6. Data Sanitization and Redaction
Before content is indexed for AI, the system should sanitize and validate extracted text.
- Remove unsupported formatting and irrelevant navigation text from scraped pages.
- Detect and redact unnecessary sensitive data where applicable.
- Validate source category and property/community assignment.
- Prevent indexing of draft or unapproved documents.
- Log sanitization and indexing status for audit.
- If content is stale, missing, or unapproved, AI should not invent answers.
7. Technical Architecture
Required planning item: Include S3 + vector search + PMC/community isolation.
flowchart LR A[Document Center UI] --> B[Metadata Service] B --> C[(Document Metadata DB)] B --> D[S3 Secure File Storage] B --> E[Approved URL Registry] D --> F[Extraction Worker] E --> G[URL Fetch + Scrape Worker] F --> H[Sanitization + Chunking] G --> H H --> I[Embeddings] I --> J[(Vector Search Index)] C --> K[Scoped Retrieval Filter] J --> K K --> L[Response Drafting] K --> M[AI Chat Widget] K --> N[Internal Knowledge Lookup]
- Every document and URL has PMC/community/property metadata.
- Retrieval queries must apply tenant and community filters before returning context.
- Deleted/archived content is removed from active retrieval.
- AI retrieval should prefer the latest active approved source.
8. AI Usage
AI response drafting
Draft responses grounded in current approved pricing, policy, and amenity documents.
AI Chat Widget
Website assistant answers common questions using property-scoped Document Center content.
Custom reports
Reports can summarize document readiness and highlight stale property knowledge.
Onboarding readiness
New communities can be checked for required documents before AI enablement.
9. User Cases
URL-based knowledge refresh
Scenario: Customer maintains amenity details on their property website.
System behavior: Approved URL is re-scraped, sanitized, and indexed.
Result: AI uses fresh website content without manual PDF upload.
Stale pricing document warning
Scenario: Pricing sheet has not been updated in 90 days.
System behavior: Document Center flags the file for review.
Result: AI can be restricted from quoting stale pricing until updated.
10. Acceptance Criteria
- User can upload files and add approved URLs.
- User can assign each source to PMC/community/property.
- User can replace, archive, delete, re-scrape, and re-index sources.
- System tracks version, active status, last updated, last indexed, and owner.
- System sanitizes content before indexing.
- System stores documents in S3 or equivalent secure storage.
- System indexes extracted text in vector search.
- AI retrieval respects PMC/community isolation.
- AI uses only active approved current sources.