nathanrenting.dev
Project · pre-development

PhotoFlow — from edited photo to online in 30 seconds

Taking out the work between "shoot edited" and "shoot online". For professional photographers working in Lightroom: a single approval delivers Instagram teasers, ceremony carousels, blog coverage, portfolio update and LinkedIn case study all at once — scheduled, published and monitored.

Hand-drawn pipeline: SHOOT → LIGHTROOM EDIT → 'send to PhotoFlow' → DESKTOP CLIENT (Tauri) → four agents (VISION/COPY/PUBLISHING/ANALYTICS) → IG, FB, LinkedIn, BLOG. Annotation '30 SECONDS' at the top, 'one approval → all deliverables' at the bottom.

Whiteboard sketch · from shoot to deliverables in 30 seconds

What the photographer still does

Three steps:

  1. Take the photo
  2. Edit the photo in Lightroom
  3. Approve

Everything in between — captions, hashtags, scheduling, publishing to IG / FB / LinkedIn, blog, portfolio — is done by AI.

Core concepts

Project / Shoot is the main object. Not "posts". A project ("Wedding Lisa & Mark") is translated by AI into multiple derivatives. One approval → all deliverables ready.

Curation vs Technical QA. AI only does the objective technical assessment: sharpness, eyes-open, exposure, motion blur. The artistic selection stays entirely with the photographer via Lightroom ratings.

Toggle-first UX. All features set up once during onboarding (four screens), after that only via Settings. Defaults aggressively ON. The workflow never adds steps to what the photographer is already used to.

RAW-blind. PhotoFlow does not read RAW. Lightroom renders; PhotoFlow operates on previews and exports. Works for Canon, Fuji and every brand supported by Lightroom without special-casing.

Architecture

A hybrid setup:

ComponentTechResponsibility
Desktop clientTauri + React + SQLiteLocal: photos, projects, drafts, settings
Lightroom pluginLua (.lrplugin)"Send to PhotoFlow" — talks to the desktop client via localhost
Cloud componentSupabase or Cloudflare Workers + D1 + R2OAuth proxy, token refresh, scheduled-publish queue, App Review endpoints

The local-first choice keeps photo data out of cloud storage unless that is necessary. The cloud only does what it has to: social-network OAuth and publish scheduling at a time when the photographer's device may be switched off.

Multi-agent

AgentFunction
VisionImage analysis, subject detection, mood
CopyCaptions, hashtags, SEO, blog
PublishingScheduling, publishing, status checks
AnalyticsResults, reporting, improvement advice

Four roles, one coordinator. The pattern behind a large part of ECHO recurs here: a thin router layer distributes the work and the output lands in one structured place.

Audience + API reality

Professional photographers with:

PhotoFlow does not support personal social-media accounts. Not a choice but a hard API limitation of Meta and LinkedIn. Better to say that up front than to discover it during onboarding.

Status

Pre-development. Vision, architecture decisions, audience validation and data model captured. Next step: scaffold