An open-source AI investment companion for Thai index investors.
Macrotide closes the gap between index investing's proven theory and how DIY investors actually execute it. It pairs education, portfolio analysis, market research, and low-fee Thai fund selection with an AI advisor grounded in your real holdings.
⚠️ Experimental — a personal-use project, not investment advice.
A responsive app built around five main tabs — Portfolio, Markets, Explore, Advisor, and Journal — across mobile / tablet / desktop with light/dark/system themes. At its core: hold your Thai mutual-fund holdings, visualize allocation, fees, and NAV trends, pull a live market view from free public sources, and chat with an AI advisor grounded in your real portfolio, plan, and journal — including plan edits it proposes as accept/reject cards.
Portfolio, ledger & analytics
- Unified ledger — your holdings are a projection of one buy/sell + balance ledger; full-screen History and per-fund Position pages, in-place edits, custom self-priced assets, and graceful handling of unknown cost basis
- Performance — headline total return on contributed capital, with a full breakdown: realized and unrealized gains, money-weighted return (XIRR) per-fund and whole-portfolio, cost-basis timeline, contributions and income
- Wealth-over-time chart — portfolio value vs. net invested over time, replayed from your ledger so it shows what you actually held (exited funds included)
- Plain-language health checks — allocation drift, blended fees, cash drag, and look-through diversification (underlying single-name concentration, feeder-aware); fee-creep alerts you can dismiss or snooze
- Flexible import — paste, a screenshot, CSV, or typed rows, all reviewed in one editable list before saving; symbol autocomplete; enter by units or baht value
- Broker auto-sync — connect a broker once and it imports your full order history and keeps it in sync, each account as its own portfolio
- Plan & journal — a markdown investment plan the Advisor proposes edits to, plus notes, decisions, questions, and a reading list
AI Advisor
- Grounded streaming chat — knows your real portfolio, plan, and journal; tool-calls, accept/reject proposal cards, performance-vs-index and plan-anchored rebalancing guidance
- In-chat vision — attach images for multi-image holdings reconciliation (snapshot vs transaction-history auto-detected) into an editable importer, plus chart / factsheet Q&A
- Long-term memory + chat archival — recall, full-text search, and session lifecycle with preference extraction; you correct the Advisor in words (no ratings bar) and it remembers, with every write shown inline and reversible
- Built to be reliable — empty-turn recovery, provider fallback, a configurable cheap-paid public tier with token / cost caps, and a committed eval harness in CI
Funds & markets
- Fund finder & screener — fuzzy, feeder-aware search over priceable share classes with per-class fee, tax wrapper, and 1-year return; screen by tracked index (S&P 500, SET50…) for the cheapest trackers; shows buyable funds by default while search finds any active fund you might hold; TER ranking
- Fund detail — per-share-class price and fund-size (AUM) history, holdings by asset type, and feeder look-through to underlying holdings
- Market data — live index levels (FMP / EODHD with ETF-proxy + Yahoo fallback) + FX, Thai fund NAVs and history (Thai SEC), and RSS news, over a resilient stale-on-error cache
- Benchmarks & models — match-or-beat the market on a total-return basis (global, US, regional, and Thai indices), and model portfolios you can browse, fork, and set as a target that drives drift + health checks
Platform
- Sign-in — passkeys + Google, with an isolated, ephemeral per-session demo mode
- Multi-user — per-user data isolation, tiers and quotas, owner admin
- Storage — a two-database SQLite split (precious
app.db+ regenerablemarket.db) on Drizzle, with daily backups - Scheduled jobs — daily NAV freshness + all-funds NAV/AUM history pre-warm on systemd timers, plus a nightly SEC fund-data ELT pipeline with risk-spectrum asset classification
For what's next see the project board; for shipped detail see CHANGELOG.md and the docs/ user + developer guide.
- Next.js 16 (App Router) + React 19 + TypeScript
- Hand-rolled CSS — design tokens, light/dark/system themes (no Tailwind)
- Hand-rolled SVG sparklines + recharts interactive charts
- Biome for lint and format; simple-git-hooks with lint-staged for pre-commit
- SQLite + Drizzle ORM for persistence; per-session in-memory SQLite for the demo mode
- Vercel AI SDK via OpenRouter for chat (one key, every major model)
- better-auth + passkeys for sign-in — see auth-and-providers.md, SECURITY.md, deploy.md, AGENTS.md
git clone <repo-url> macrotide
cd macrotide
npm install
npm run devOpen http://localhost:3000. A fresh boot lands on /login; click
Try the demo to spin up an isolated in-memory SQLite seeded with mock
data (capped at 10 chat turns). For solo localhost dev, copy .env.example
to .env.local and set AUTH_DISABLED=1 to skip the login screen — see
auth-and-providers.md. Chat returns a friendly stub until you set
OPENROUTER_API_KEY.
Scripts:
npm run dev # dev server (hot reload)
npm run build # production build
npm run start # serve production build
npm run lint # Biome check
npm run format # Biome check --write
npm run typecheck # tsc --noEmitmacrotide/
├── app/
│ ├── (auth)/login/ Passkey sign-in screen
│ ├── api/ Route handlers: buckets, holdings, journal, plan,
│ │ models, quotes, settings, chat (+threads),
│ │ market, demo, auth/[...all], admin
│ ├── layout.tsx, page.tsx, error.tsx, globals.css
├── components/
│ ├── screens/ Portfolio, Markets, Advisor (chat), Journal,
│ │ Models, Connect, Settings (Explore lives in FundSelect.tsx)
│ ├── App.tsx, ClientApp.tsx, AppPanels.tsx, charts.tsx, *Sheet.tsx, …
├── lib/
│ ├── ai/ OpenRouter provider + chat plumbing
│ ├── api/ Rate-limit + with-db helpers for route handlers
│ ├── auth/ better-auth singleton + session helpers
│ ├── db/ Drizzle client, schema, migrations, queries,
│ │ per-session demo SQLite, daily backup
│ ├── fetchers/ SWR fetchers (client-side data layer)
│ ├── market/ Provider registry + cache + indices
│ ├── mock/ Seed data + demo seed (used by db:seed)
│ ├── portfolio/ Allocation/concentration analytics, plan parser,
│ │ plan-edit helper
│ ├── static/ Editorial content (markets/learn/personalities)
│ │ and placeholder analytics
│ ├── format.ts, useViewport.ts, useScrollHide.ts
├── packages/
│ └── connector-sdk/ @macrotide/connector-sdk — broker-agnostic
│ connector contract + parser/collector builder
├── data/ SQLite + daily backups (gitignored)
├── tests/ Vitest
├── docs/ User + developer guide (Diátaxis)
│ ├── tutorials/ Learning by doing (getting-started)
│ ├── how-to/ Tasks (local-development, import, deploy)
│ ├── reference/ Lookup (configuration, auth-and-providers, api, data-model, design-system)
│ └── explanation/ Why (architecture, design-principles, memory, decisions/ + research/)
├── llms.txt Machine-readable docs map for AI agents
├── README.md, AGENTS.md, SECURITY.md, LICENSE
Only convention-mandated files stay at the repo root (README, AGENTS, SECURITY,
LICENSE); everything else navigable lives under docs/.
Full docs live in docs/, organized with the Diátaxis framework:
- New here? docs/tutorials/getting-started.md
- Building on it? docs/how-to/local-development.md and docs/explanation/architecture.md
- Looking something up? docs/reference/ (config, API, data model)
- An AI agent? Read AGENTS.md, then llms.txt → docs/README.md.
This is an experimental personal project. PRs and issues are welcome but expect slow / opinionated responses. If you're picking up something from the project board, comment on the issue first so we don't duplicate work.