Gemini via Vertex AI
Planning and drafting from a confirmed brief. Start with one editable content draft, then evaluate quality, factual consistency, latency and cost.
Official documentation ↗TECHNOLOGY & PRODUCT STATUS
A shared practice brief, one coordinated plan and eight specialist roles. Below is what works today, what is being built next, and our proposed Google Cloud architecture.
The public site currently runs on Cloudflare Pages, Workers and D1. Google Places is the live Google API integration. The preview makes no live language-model requests.
| Capability | Status | Implementation |
|---|---|---|
| Public clinic lookup | Live | Google Places API through a server-side Cloudflare Worker; practice details, photos and Maps links. |
| Demo & contact requests | Live | Cloudflare Worker + D1: validated requests, consent, reference IDs and rate limits. |
| Workspace, plans & workrooms | Preview | Static HTML / CSS / JavaScript. Local templates, guided setup, project dependencies and draft review. |
| AI planning & website/content generation | In development | Primary planned path: Google Gemini models via Vertex AI; no live model calls in the preview. |
| Authenticated Google / Meta connections | In development | Setup flow preview only. Owner authorization, token storage and account scoping must be implemented. |
| Voice reception & booking | In development | Optional voice adapter plus an authorized phone and booking provider; no calls or bookings today. |
| Publishing, social posting & ad spend | In development | Future separate execution approval, provider permissions, budget limits and audit records. |
PLANNED ARCHITECTURE ON GOOGLE CLOUD
We intend to use Google Cloud for the first live generation workflow and an authenticated pilot backend. These services are proposed, except the existing Google Places integration. We do not claim Google Cloud program approval, a partnership or endorsement.
Planning and drafting from a confirmed brief. Start with one editable content draft, then evaluate quality, factual consistency, latency and cost.
Official documentation ↗A containerized backend for authenticated requests, orchestration and provider adapters. Keep each execution scoped to the practice and approved task.
Official documentation ↗Proposed storage for practice briefs, plans and approval records. Firestore is the initial design choice; final data model and region selection are pending.
Official documentation ↗Planned storage for generated images, documents and media, with access controls and a defined retention policy.
Official documentation ↗Google Places clinic discovery already works today. Public lookup runs through our existing Cloudflare backend, not a Cloud Run deployment.
Official documentation ↗Planned credential isolation, least-privilege service identities and observability with sensitive input redaction.
Official documentation ↗The first milestone is a confirmed practice brief → Gemini content draft → human editing and approval. Credits would support development, quality evaluation and small pilot workloads before broader orchestration.
Our planned adapters accept a scoped task and approved practice facts, then return a structured draft for review. Google Gemini is the primary model path. A provider can be replaced without changing the practice brief or approval flow. This adapter layer is a design direction; it is not yet a live multi-model runtime.
Google / Gemini architecture ↗Practice details are confirmed before drafting. The planned model workflow will identify missing facts instead of filling them with unsupported claims.
Draft approval records a review decision today. Publishing, spending, calls and booking are disabled and will need separate authorization when implemented.
The preview has a local activity feed. Authenticated audit records, access controls, recording consent and production retention rules are planned.
Start with a free clinic lookup or explore the public workspace.