CutTrack AI — standalone version

An Indian calorie and macro tracker that runs on your iPhone's Home Screen. It does not depend on a Claude plan.

iPhone (CutTrack app)
   │  your log is kept on the phone first, so it works offline
   ▼
Your Vercel server (free)  ── holds your secrets ──┐
   ├─ /api/drive       →  Google Drive   (stores your data, no AI)
   ├─ /api/ai          →  OpenAI → Gemini → Groq   (reads meals, photos, coach)
   └─ /api/transcribe  →  OpenAI → Gemini → Groq   (voice notes)
                          if one runs out of credit or fails, the next takes over

What lands in your Drive

My Drive/
└── CutTrack AI/
    ├── cuttrack-data.json        ← everything, used to restore on a new phone
    └── 2026/
        └── 10 - October/
            └── 2026-10-18.txt    ← that day's entries with time, amount, macros and totals

CutTrack can only see files it created. It cannot read anything else in your Drive.

Cost: Vercel Hobby and Google Drive are free. OpenAI is pay-as-you-go; gpt-6-luna costs $0.10 per million input tokens, so a month of logging typically costs a few rupees. Gemini and Groq take over only when OpenAI can't answer.

---

Before you start

You need:

• a laptop (the setup is far easier there than on a phone)
• a GitHub account
• a Google account ([email protected])
• about 30 minutes

AI keys live in GitHub secrets. The running app on Vercel can't read GitHub secrets directly, so the included Sync AI keys to Vercel action copies them across. It needs one extra secret, VERCEL_TOKEN (step 4b).

---

Step 1 — AI keys (already done)

OPENAI_API_KEY and GEMINI_API_KEY are in the repository's GitHub secrets. Add GROQ_API_KEY there whenever you have it, then run the sync action (step 4b).

OpenAI bills API use separately from ChatGPT: add credit at platform.openai.com → Billing.

Step 2 — Put the code on GitHub

1. Unzip cuttrack-app.zip on your laptop.
2. On github.com, click New repository, name it cuttrack-ai, choose Private, then Create repository.
3. On the empty repo page, click uploading an existing file.
4. Drag everything inside the unzipped cuttrack-app folder into the page. That means the api and icons folders plus index.html, sw.js, manifest.webmanifest, package.json, vercel.json and README.md.
5. Click Commit changes.

Check that the repo shows an api folder at the top level, not a cuttrack-app folder containing it.

Step 3 — Set up Google sign-in (Google Cloud)

1. Go to console.cloud.google.com. Accept the terms, then create a project named CutTrack. Make sure it's selected at the top.

2. Turn on the Drive API. Search "Google Drive API" in the top bar, open it, and click Enable.

3. Set up the sign-in screen. Open Google Auth Platform; in older layouts it's APIs & Services → OAuth consent screen.

   • Click Get started.
   • App name: CutTrack AI. Support email: your Gmail address.
   • Audience: choose External.
   • Contact email: your Gmail address. Agree to the terms, then click Create.

4. Choose what the app may access. Open Data Access → Add or remove scopes. Tick:

   • .../auth/drive.file
   • openid
   • .../auth/userinfo.email

   Click Update, then Save.

5. Publish the app. Open Audience → Publish app → Confirm. This matters: if the app stays in "Testing", Google signs you out every 7 days. The scopes above are non-sensitive, so no Google review is needed.

6. Create the sign-in credentials. Open Clients; in older layouts it's Credentials → Create credentials → OAuth client ID.

   • Application type: Web application. Name: CutTrack.
   • Leave Authorized redirect URIs empty for now; you'll add it in step 5.
   • Click Create.
   • Copy the Client ID and Client secret into your note.

Step 4 — Deploy on Vercel

1. Go to vercel.com and sign up with GitHub.

2. Click Add New… → Project, then Import your cuttrack-ai repo.

3. For Framework Preset, choose Other. Leave build settings empty.

4. Open Environment Variables and add each of these:

   Name	Value
   GOOGLE_CLIENT_ID	from step 3.6
   GOOGLE_CLIENT_SECRET	from step 3.6
   ALLOWED_EMAIL	[email protected]
   SESSION_SECRET	a random string of 32 or more characters (see below)

   For SESSION_SECRET, use a password generator, or run openssl rand -base64 36 on a Mac. Don't reuse a real password.

5. Click Deploy. When it finishes, copy your address, for example https://cuttrack-ai-xyz.vercel.app.

Step 4b — Let GitHub pass the AI keys to Vercel

1. Go to vercel.com/account/tokens → Create Token. Name it cuttrack-github, scope it to your account, choose an expiry, and copy it.
2. On GitHub: cuttrack-ai → Settings → Secrets and variables → Actions → New repository secret. Name VERCEL_TOKEN, paste the token, then click Add secret.
3. Open the Actions tab → Sync AI keys to Vercel → Run workflow → Run workflow.
4. When it shows a green tick, the keys are in Vercel and the app has redeployed.

Run it again whenever you add or change an AI key, for example when you add GROQ_API_KEY.

Step 5 — Connect the two

1. In Vercel: go to Settings → Environment Variables and add APP_URL with your address. No slash at the end.
2. Then go to Deployments, open the ⋯ menu on the latest deployment, and choose Redeploy.
3. In Google Cloud: go to Clients → CutTrack, then Authorized redirect URIs → Add URI. Enter:

   https://YOUR-ADDRESS.vercel.app/api/auth/callback
   

   Then click Save. Google can take a few minutes to apply it.

This is the beginning of the README. Read the rest at the original. Read the full README ↗