Skip to content

distractedhare/customerconnect-ai

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

577 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CustomerConnect AI

iPhone-first coaching for T-Mobile virtual retail reps.

The maintained app is now the single iOS app in apps/ios. Its primary runtime is the built React/Vite PWA bundled into the app, so the iPhone app keeps the cyberpunk Kip visuals, liquid/glass styling, scroll behavior, and current rep workflows while the hosted PWA winds down.

There should be exactly one Xcode project for this product:

apps/ios/CustomerConnectiOS.xcodeproj

Native iOS App

Open the team project:

npm run ios:open

Build the app for the simulator:

npm run ios:build

Run the full local iOS doctor check:

npm run ios:doctor

Run the core Swift tests:

npm run ios:test

Run the full iOS guardrail:

npm run guard:ios

Canonical Xcode project:

apps/ios/CustomerConnectiOS.xcodeproj

Canonical scheme:

CustomerConnectiOS

Use CustomerConnectiOS when Xcode Cloud asks which product to build. CustomerConnectCore is the shared Swift package used by the app; it is not the installable app.

Legacy PWA

The PWA source remains because it is the source of truth for the iOS app's bundled web runtime.

  1. Install dependencies: npm install
  2. Start the app: npm run dev

The app always ships with instant local coaching from the embedded data layer plus the current weekly-update.json.

Optional AI assist

AI enrichment is optional and never blocks the rep flow. The PWA builds a local plan first, then tries to sharpen it in the background when AI is configured.

Kip chat uses deterministic app/generated data as the factual answer first. Optional local model polish can run in-browser through MediaPipe LLM Inference with Gemma 3 270M IT web:

  • Default model path: /models/gemma3-270m-it-q8-web.task
  • Override path: VITE_KIP_GEMMA_MODEL_URL
  • If the local model pack is missing or unsupported, Kip still answers from local app knowledge.

Recommended for deployment:

  • AI_COMPLETIONS_URL
  • AI_MODEL
  • AI_API_KEY

Those power the Vercel /api/ai proxy so keys stay off the client.

Direct browser fallback is also supported for local/internal testing only:

  • VITE_AI_COMPLETIONS_URL
  • VITE_AI_MODEL
  • VITE_AI_API_KEY

Verify Gemma before launch

Health check only:

npm run check:ai -- --url=https://meridianvrtesting.com --health-only

Full smoke test:

npm run check:ai -- --url=https://meridianvrtesting.com

The smoke test fails fast if the Vercel proxy is missing config, the auth is wrong, or the model returns an empty response.

Monday-safe behavior

  • No AI config: the app still works with local coaching only.
  • AI timeout or provider failure: the local plan stays visible.
  • First load: the app now waits for the weekly data before generating the plan so promos and intel populate consistently.

About

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors