spl1t is an open source expense-tracking app based on Spliit. This fork deploys on Cloudflare Workers (via OpenNext) with Cloudflare KV as the database — not Vercel Postgres / Prisma.

Live: https://spl1t.pages.dev

spl1t.pages.dev is a Pages project that reverse-proxies Worker spl1t. OpenNext still deploys the Worker. App NEXT_PUBLIC_BASE_URL remains https://spl1t.USER.workers.dev until a Worker rebuild.

Features

Legend: 🟢 from original Spliit · 🔴 new in this Cloudflare KV fork

• 🟢 Create a group and share it with friends
• 🟢 Create expenses with description
• 🟢 Display group balances
• 🟢 Create reimbursement expenses
• 🟢 Progressive Web App
• 🟢 Select all/no participant for expenses
• 🟢 Split expenses unevenly
• 🟢 Mark a group as favorite
• 🟢 Tell the application who you are when opening a group
• 🟢 Assign a category to expenses
• 🟢 Search for expenses in a group
• 🟢 Export a group to JSON or CSV
• 🔴 Import a group from a Spliit JSON export (creates a new group with remapped IDs)
• 🔴 Import a group from a Tricount GDPR CSV export (participants + expenses)
• 🔴 Notes + activity history + document links in JSON export/import (exportVersion: 3)
• 🔴 Copy an existing expense into a new draft
• 🔴 Math expressions in the amount field
• 🔴 Even-split cent remainder (no missing cents)
• 🔴 Group default split mode
• 🔴 Optional group PIN
• 🔴 Share group via QR code
• 🔴 Soft-delete / restore groups (30-day grace) + 24-month inactivity expiry
• 🔴 Security headers, CSV formula escape, Zod input caps, expense date bounds
• 🔴 Error boundaries + Drawer accessibility titles
• 🔴 Paste-friendly amount parsing (US/EU grouped currency)
• 🔴 Keyboard navigation restored in category/currency selectors
• 🔴 Mobile group tab icons
• 🔴 Stats: monthly spending (stacked category chart + breakdown) and balance timeline
• 🔴 Optional calendar-month expense grouping
• 🔴 Drag-reorder / Sort A–Z participants
• 🔴 Multiple payers per expense (legacy single paidById migrated on read)
• 🔴 Extra currencies: ARS, TRY, COP, JOD, MKD, MOP, MYR, VND
• ❌ Upload and attach images to expenses (removed — see below)
• ❌ Create expense by scanning a receipt (removed — see below)

Stack

• Next.js for the web application
• TailwindCSS for the styling
• shadcn/UI for the UI components
• Cloudflare KV for persistence (denormalized group documents)
• OpenNext Cloudflare + Workers for hosting

Data model notes

• Each group is stored as a single KV value under group:{groupId}.
• Categories are seeded under the categories key.
• Concurrent edits to the same group use last-write-wins (no Durable Objects / transactions).
• Friend-sized groups fit this model; very large groups may hit KV value size limits.
• Groups track lastActivityAt on expense create/update/delete and group settings updates. After 24 months without activity, cleanup soft-deletes them; soft-deleted groups can be restored for 30 days, then are hard-deleted. Call GET/POST /api/cron/cleanup-groups with Authorization: Bearer $CRON_SECRET (set CRON_SECRET in Worker env).

Extra UX (this fork)

Ideas below track community demand from Spliit Cloud’s roadmap, upstream Spliit issues/PRs, and hardening patterns from anon-spliit (reimplemented for KV — not a code port of their E2EE/auth stack).

Feature	Notes	Prior art
Copy expense	From the expense list or edit header icon (opens create prefilled).	Upstream #527; shipped in Spliit Cloud
Amount math	Expressions in the amount field (10+5.50, 5*8, …) on blur/save.	Upstream #184; shipped in Spliit Cloud
Default split mode	Stored on the group (device localStorage can still override).	Upstream #366; shipped in Spliit Cloud
Even-split cents	Integer remainder allocation so balances don’t drop a cent.	Upstream #374 / #427; tracked by Spliit Cloud
Share QR	QR in the share popover.	Upstream #500; on Spliit Cloud roadmap
Optional group PIN	4–8 digits; unlocks per browser session; hashed on the server, not returned to clients.	Upstream #373; on Spliit Cloud roadmap
Notes + history + document links in JSON	Export/import round-trips expense notes, group information, activity history, and document URLs (exportVersion: 3).	Follow-up to upstream #546; expense notes also in #165
Soft-delete + inactivity expiry	Manual soft-delete with 30-day restore; auto soft-delete after 24 months without activity; cron hard-deletes after grace.	Inspired by anon-spliit deletion/auto-delete work and upstream #420
Paste amount parsing	Normalizes pasted US/EU currency amounts in number fields.	Upstream #531
Selector keyboard nav	Category/currency pickers use cmdk CommandList.	Upstream #491
Mobile tab icons	Icon-only tabs on small screens; labels from sm.	Upstream #539
Monthly spending + balance timeline	CSS stacked category charts, category breakdown, and balance timeline on Stats.	Upstream #532 / #555
Calendar month grouping	Optional group setting for roommate-style monthly lists.	Upstream #530
Multiple payers	Split who paid an expense across several participants; balances/export/import aware. Legacy paidById migrates on read.	Upstream #396
Reorder participants	Drag-and-drop + Sort A–Z; order persisted in KV.	Upstream #416
Tricount import	GDPR CSV export via the same Import control as Spliit JSON.	Upstream #526
Export / input hardening	CSV formula escape, Zod max caps, expense date bounds, security headers, error boundaries.	Patterns reviewed from anon-spliit (adapted for Workers/KV)

Stats (this fork)

On each group’s Stats tab:

• Monthly spending — stacked category chart for calendar months, with a category breakdown and legend controls.
• Balance timeline — cumulative balances over time for participants (engineering fixes on this fork for share math / timeline consistency).

Inspired by upstream #532 / #555; reimplemented for denormalized KV documents.

Group import JSON / Tricount (this fork)

On the Groups page, use Import JSON to upload:

1. A Spliit JSON export (this fork or upstream Spliit), or
2. A Tricount personal-data / GDPR CSV export (detected automatically).

Shared behavior:

• Always creates a new group (does not overwrite an existing one).
• Regenerates group, participant, and expense IDs so imports never collide with live data.
• Does not import a group PIN (PIN must be set again after import).

Spliit JSON

• Restores participants, expenses (including multiple payers when present), split modes, amounts, dates, notes, group information, and activity history (when present in the file).
• Categories: match by id when present; otherwise by name / grouping against the seeded list (many exports omit id).
• Newer exports include exportVersion: 3, expense ids (needed to re-link history), and expense document links (url / dimensions — not file bytes).
• Document URLs round-trip for migration between Spliit forks; binaries are not embedded, and links may 404 if the original storage expires. Recurring-expense links are still not restored.

Tricount CSV

• Imports participants and expenses (amounts by share / impacted amounts).
• Uses the CSV default currency (Frankfurter .dev for missing cross-rates).
• Notes and activity history are Spliit-only; Tricount imports leave them empty.
• Prior art: upstream #526.

Removed / disabled upstream features (S3 & OpenAI)

Upstream Spliit optional features that depended on AWS S3 and OpenAI are not available in this Cloudflare KV fork:

Feature	Upstream dependency	Status here
Expense document / image uploads	S3 (or compatible object storage)	Removed from the critical path; UI/API stubs keep flags off. KV is not used for binaries.
Create expense from receipt scan	OpenAI + storage	Disabled; no OpenAI client or API keys.
Category extract from text/image	OpenAI	Disabled; same as above.

What changed vs upstream:

• Prisma, Postgres, and Vercel-oriented DB wiring were replaced with the KV group-document API.
• S3/OpenAI packages and env vars were dropped; keep NEXT_PUBLIC_ENABLE_EXPENSE_DOCUMENTS, NEXT_PUBLIC_ENABLE_RECEIPT_EXTRACT, and NEXT_PUBLIC_ENABLE_CATEGORY_EXTRACT unset or false (see .env.example).
• Re-enabling uploads later would mean adding something like R2 (not stuffing files into KV). Receipt/category AI would need a Workers-compatible provider and explicit product work.

Run locally

1. Clone the repository: git clone https://github.com/t0ma5/spl1t.git
2. Copy .env.example to .env and .dev.vars as needed
3. Create a KV namespace and put its id in wrangler.jsonc:

npx wrangler kv namespace create spl1t-db
npx wrangler kv namespace create spl1t-db --preview

4. Run npm install (uses package-lock.json)
5. Set NEXT_PUBLIC_BASE_URL (production default in wrangler.jsonc vars is https://spl1t.pages.dev)
6. Run npm run dev for Next.js local development (bindings via OpenNext), or npm run preview to build and run in the Workers runtime

Note: Local OpenNext/Wrangler needs workerd, which does not support Windows ARM64. On those machines, develop against the remote Worker or deploy from an x64/Linux host.

Deploy to Cloudflare

Deploy directly to Cloudflare (no GitHub Actions).

Requires Node.js 22+ and a host where Wrangler/workerd runs (Linux / macOS / Windows x64 — not Windows ARM64).

npm run deploy

This runs opennextjs-cloudflare build then deploys Worker spl1t. Ensure:

• DB KV binding in wrangler.jsonc points at your namespace (existing id kept so group data survives).
• vars.NEXT_PUBLIC_BASE_URL matches the URL users open (https://spl1t.pages.dev).

Ops notes

• Pushing code to GitHub does not update the live Worker until you run npm run deploy (or equivalent OpenNext/Wrangler upload) against Cloudflare.
• Prefer git / GitHub CLI over the GitHub web “upload files” UI — uploads often drop directories.
• Set Worker secret CRON_SECRET and schedule a daily call to /api/cron/cleanup-groups for inactivity cleanup.

Health check

• GET /api/health/readiness or GET /api/health — app ready, including KV connectivity
• GET /api/health/liveness — process alive only

Credits & provenance

• Original Spliit — idea, UI, and core expense-splitting product by Sebastien Castiel and contributors: spliit-app/spliit · spliit.app.
• Spliit Cloud (antonio-ivanovski/spliit-cloud) — community fork that continues Spliit with new features. Several UX improvements in this Workers/KV fork were prioritized from their roadmap and upstream issue links (reimplemented for denormalized KV documents, not a code port of their Postgres/API stack).
• anon-spliit (sora-grayscale) — privacy-focused fork (E2EE, private instance, deletion/auto-delete). This Workers/KV fork adapted selected lifecycle and hardening ideas from that work; it does not port their end-to-end encryption or account/2FA stack.

License

MIT, see LICENSE. Same license family as upstream Spliit and Spliit Cloud; retain their copyright notices where applicable.