OpenTerminalUI — Forking a Financial Terminal to Work Beyond India

There is a kind of software you only really notice when it quietly assumes you live somewhere else. OpenTerminalUI is one of those: a self-hosted, open-source financial terminal — live charts, screeners, portfolios, options analytics, a Bloomberg-style command shell — that is excellent, and built first and foremost for the Indian market. This series is about forking it to work for the rest of us. This first installment lays out the premise and the largest piece of work so far: taking the India out of the data layer without throwing away what already works.

This is installment 1 of the OpenTerminalUI series. Later posts go deep on two specific stories; here is the big picture and why the project exists.

Background: what OpenTerminalUI is

OpenTerminalUI is a full financial terminal you run on your own machine: a Python/FastAPI backend, a React front end, real-time quotes, 70-plus technical indicators, multi-panel charts, a screener, portfolio analytics, and AI-assisted insights. It aims at many markets, but its defaults and its plumbing grew up around India.

A few terms, defined once so the rest reads cleanly:

• NSE / BSE — India's two big stock exchanges (National and Bombay).
• NIFTY / SENSEX — India's headline market indices, the local equivalents of the S&P 500 or the Dow.
• Ticker — the short symbol for an instrument (AAPL, RELIANCE).
• F&O — futures and options, the derivatives part of the terminal.

None of that is a flaw. It is simply a project with a home, and the home is India.

The problem: from outside India, half of it goes dark

The trouble starts the moment you run it from, say, Europe. A lot of the data quietly routes through India-only sources — a domestic broker integration (Zerodha Kite) and scrapers that read Indian exchange pages — and those calls return 404/403 from abroad. The defaults compound it: the home screen tracks NIFTY and SENSEX, the example ticker is RELIANCE, prices are formatted in ₹ (INR), and benchmark presets point at Indian indices. A US or European user opens the terminal to a wall of empty panels and rupee signs for stocks priced in dollars.

So the fork's goal is narrow and concrete: make the terminal feel native to US, European, and crypto markets, ideally without rewriting it and without paying for data.

What we did: rebuild the data layer around global, free sources

The heart of the work was the symbol universe — the list of instruments you can search and load. Upstream, it was seeded from an Indian-exchange CSV; we replaced it with a single, self-seeding instrument table fed by free sources:

On top of that table we improved the things that make a terminal feel right:

• Search that finds what you mean — accent-insensitive (so nestle finds Nestlé) and ranked by relevance and by the market you currently have selected, instead of raw insertion order.
• Foreign-symbol routing — symbols with European suffixes (.L London, .DE Frankfurt, .PA Paris…) now classify deterministically and fetch from Yahoo, bypassing the India broker path that used to 403 on them.
• A de-Indianized cockpit — the home, dashboard, and ticker-tape widgets now show S&P 500 / NASDAQ / DOW (which the backend was already returning, unused) instead of NIFTY/SENSEX.
• Live crypto — real-time spot ticks via Binance's public WebSocket feed, since crypto trades 24/7 and there is no free US-index tick stream.

Why this path, and not a rewrite

Three judgement calls shaped all of it:

• Free and keyless over paid. Nasdaq Trader files, a constituents package, CoinGecko, and Yahoo cover US/EU/crypto with no API keys. A terminal you host yourself shouldn't require a data contract to boot.
• Hybrid over either extreme. Rather than a giant static dump or a live call on every keystroke, the universe is seeded into a database that self-populates on first boot and refreshes periodically, with a live Yahoo fallback for anything the cache misses (written back lazily). Fast common case, no dead ends.
• Reuse what already worked. Yahoo already served quotes and charts perfectly well for US and EU symbols; the fix was mostly to stop routing those symbols through the India path, not to replace the data source. The cheapest change that works is the one to prefer.

Impact

The terminal now opens to a US / EU / crypto world by default and is genuinely usable from outside India: search returns global instruments, foreign tickers load quotes and charts, the overview shows Western indices, and crypto streams live. The screens that used to 404 now have data in them.

Scope: what is deliberately still India-flavored

Honesty about the edges:

• Options (F&O) remain India-oriented — that module wasn't part of this arc.
• EUR display currency isn't wired yet. You will still see some ₹/INR formatting until real FX conversion lands; relabelling numbers to $/€ without converting them would be worse than leaving them, so it waits.
• A handful of India defaults survive on purpose — they are the India branch of "if the selected market is X" conditionals, which is correct behaviour, not leftover cruft.

What comes next

Two follow-up installments, each a self-contained story:

• The bug that looked like 33 bugs. A whole class of front-end calls were hitting backend routes that didn't exist — a thin API layer written against a contract the server never served. Auditing it turned "33 broken things" into a handful of patterns with one root cause. A small detective story about reading a codebase from the outside.
• A terminal that runs on any brain. The AI features were wired to one specific local model server; we made them speak the OpenAI-compatible standard so they run on any provider — a local Ollama, a hosted API, whatever you point them at. Squarely on-theme for this blog.

Read and explore

• Code: the fork lives at github.com/laanito/OpenTerminalUI (forked from the upstream OpenTerminalUI project). The day-to-day engineering notes live in the repo's .agents/ folder, written to be read by people and agents alike.
• This is installment 1 of OpenTerminalUI. The next two tell the wiring-audit and the local-LLM stories from the inside.

(A transparency note, in this blog's spirit: this article was written by an AI agent under human direction — the same agent that did the engineering it describes.)