Wiki Authoring Platform — Instance Knowledge Base

Status: intake (brief drafted 2026-07-09) · Visibility: public Origin: 2026-07-09 UWW funding discussion (Scott Brady + Cynthia Bee, Utah Water Ways) — agreed as a new funding ask. Commercial terms of the UWW engagement are deliberately not documented in this public repo. Related direction: generative-ai-chat-interface — 2026-07-09 Direction Update

Overview

A wiki / knowledge-base platform whose content is "part of the brain that drives the AI experience" (CEO phrasing). Partner experts author and maintain the knowledge; the platform serves it three ways — to homeowners inside the guided design flow, to AI (Milo) as grounding corpus, and to the public web as education/SEO pages. One content model, three surfaces.

Why now

The 2026-07-09 direction update promoted the conversational, question-driven guide to the primary homeowner flow and added UWW extension scope: video embeds at decision points, a "before you begin" prep checklist, and step-level design-intent interview questions. All of that needs a content backbone that partner subject-matter experts can author and maintain without SimplyScapes engineering in the loop. Cynthia Bee's LocalScapes education library (classes, guides, videos — inventoried in the LocalScapes five-step-process research) is the archetype content set; the wiki is where that expertise lives, stays current, and feeds the product.

The three surfaces

1. Admin authoring interface (partner authors)

An authoring back-office through which external partner authors — water-district education staff; UWW's Cynthia Bee is the archetype author — create and maintain wiki content for their instance.

This implies a partner-author permission model: authoring rights scoped to their instance's wiki and nothing else — no access to designs, customer data, other instances, or platform administration. Today's roles are workspace-membership-based; a partner-author role is a new, deliberately narrow grant. Permission design is a key open question (see Open questions).

2. Reader side panel (Jobs tab)

A side panel in the Jobs tab that surfaces contextually relevant articles during the guided flow — the homeowner reaches a decision point (park strip layout, irrigation choice, plant selection) and the panel offers the matching article/video from their instance's wiki. This is the reader face of the candidate architecture recorded in the guided-flow brief (instance wiki in a Jobs-tab side panel).

3. Dedicated public-facing wiki pages

Public wiki pages — e.g. on the Slow the Flow portal — serving the same articles as standalone education content for SEO and program outreach. Whether these pages live on the portal domain or the app is an open question.

Content model

• Instance-scoped. A Slow the Flow wiki is distinct from any other instance's wiki; content, authors, and readers resolve per instance (same resolution pattern as the shipped whitelabel instance system).
• SSFM (GitHub-flavored markdown) as the source of truth, per the platform rich-text decision — see Agent-Native Rich Text Format (SSFM) research. Authoring uses the Plate v54 editor line; structured widgets are fenced ss: blocks. Markdown source of truth is what makes the content directly agent-readable (no serializer between the wiki and the AI).
• Media/video embeds are first-class. Instructional videos at decision points are core content, not attachments — capacity first, content later, per the guided-flow direction update.

AI ingestion — the wiki feeds Milo

Wiki content is ingested into Milo's corpus/brain — see M.I.L.O. — My Intelligent Landscape Operator and its corpus pipeline (spec — Phases: Phase 1 shipped with the whole help corpus in the system prompt; the instance wiki becomes an additional, instance-scoped corpus source). The conversational guided experience then answers from partner-authored content: when the guide explains why a park strip needs a minimum planting width, the answer grounds in the district's own wiki article. This makes the wiki the content backbone of the guided walkthrough extensions specced in generative-ai-chat-interface.

Strategic fit

• Guided flow (primary homeowner flow): the wiki supplies the videos, explanations, and jurisdiction knowledge the walkthrough surfaces at decision points.
• Whitelabel/institutional channel: the wiki is the content layer of the instance system — each district/partner brings its own expertise; the Slow the Flow portal is the first deployment target.
• Agent-native platform: SSFM markdown content is readable by Milo and any future agent without translation — consistent with the platform's markdown-as-source-of-truth direction.

Out of scope (v1 framing)

• The SimplyScapes help site (help.simplyscapes.com — the current Milo CS corpus). That content is SimplyScapes-authored product help; the wiki is partner-authored domain knowledge. They share ingestion patterns but are distinct corpora and surfaces.
• A general-purpose CMS for marketing sites — this is a knowledge base with an AI-ingestion contract, not a page builder.

Open questions

1. Partner-author roles, moderation, and review workflow. Does partner content publish directly, or through a draft → review → publish flow? Who reviews — the instance owner, SimplyScapes, or both? How is inaccurate or off-brand content handled after publish?
2. Versioning and publish states. Draft/published/archived states, version history per article, rollback — and what the AI ingests (published only, presumably) vs. what authors see.
3. Wiki-article ↔ walkthrough-step linking mechanism. How does a guided-flow decision point know which articles are relevant — explicit pinning by the author, tags/taxonomy, or retrieval (embedding search) at runtime? This contract shapes both the authoring UI and the side-panel reader.
4. Where public pages live. Portal domain (e.g. the Slow the Flow portal) vs. the app — affects SEO ownership, theming, and routing.
5. Permission model mechanics. How the instance-scoped partner-author role composes with existing workspace-membership access control (RLS) without widening any other grant.

Version History

| Date & Time (MT) | Author | Summary | |------------------------|--------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| | 2026-07-09 06:50 PM MT | Dan | Initial intake brief from the 2026-07-09 UWW funding discussion: three surfaces (partner-author admin interface, Jobs-tab reader side panel, public wiki pages), instance-scoped SSFM markdown content model with first-class video embeds, Milo corpus ingestion as the AI backbone of the guided walkthrough. Open questions recorded (partner-author permissions/moderation, versioning/publish states, article↔step linking, public-page domain). Public-safe — no commercial terms. |