538 stars · MIT — read from /blob/main/LICENSE, plain unmodified text, 'Copyright (c) 2020 nardew' · 2.7.0 of 2025-09-09 — THE YEAR TRAP: the GitHub page shows a bare '09 Sep', which the old past-date heuristic would have read as 2026; PyPI settled it as 2025 · Track this in Scout
Financial indicators for Python that recompute incrementally as each new value arrives.
▶Repo detailsthe review · specs · pros & cons · install
What it is
A Python library at 538 stars with no outside dependencies, offering the usual technical indicators in a form where adding, changing or removing a single value costs the same no matter how long the history is. Indicators can also feed each other, so one can be built on the output of another.
What it is good for. Anyone feeding live or streaming prices into indicators, where recalculating the whole series on every new value is what makes the program too slow. It is the wrong shape for working through a large stored history in one go, and the library's own design says so.
- MIT, read from the licence file and plain.
- The updating design is genuinely unusual. Nearly every other Python library of this kind works through a whole table at once, so there is no like-for-like alternative with this property.
- No outside library, no compiler and no separate C library to install first, which is exactly the problem the best-known alternative has.
- No new code for about twelve months. Version 2.7.0 and the last code change both date from 9 September 2025, and this needed proving: the release page shows the date without a year, and read carelessly it would have looked three weeks old. This is not a stale release date over live code — the code stopped too. It is well inside the eighteen-month line this report uses, and it is explicitly not being called dead, but it is the quietest project on this page.
- One maintainer is named as the sole copyright holder. There are 23 open issues and 8 open pull requests, and with no pushes in a year those contributions are simply waiting.
- The stated Python range runs from 3.8 to 3.11. There is no published statement about 3.12 or later, and nobody has pushed code to add one.
- Far fewer indicators than the mainstream libraries. You would adopt this for the updating behaviour and accept a smaller catalogue for it.
TA-Lib/ta-lib-pythonThe same job at much larger scale with more than 150 indicators and candlestick patterns, but it works through whole arrays at once and wraps a C library that has to be present.
Track this in Scout- bukosabino/ta
Pure Python and installable in one command, with 43 indicators, but it works over whole table columns for preparing data rather than accepting one new value at a time.
Track this in Scout
python3 -m venv venv && source venv/bin/activate pip install talipp
