A script that imports all historical OHLC data from the Bitfinex API and stores it in a local database.
The data has a 1-minute interval and can be used to carry out further in-depth analysis of market trends or backtest trading bots.
All currently traded symbols (e.g. BTC-USD, ETH-USD, ...) are supported (check symbols.json for a list of them). The import will begin from the earliest trading date (defined in symbols_trading_start_days.json). The script can be invoked periodically (for example with a cronjob) to fetch the latest data. It will automatically resume from the latest saved date.
API rate limits are respected and will not be exceeded. The script will re-try fetching with an incremental backoff time and continue to the next symbol after 3 failed attempts.
Experimental support for the Websocket API is also available. It currently subscribes to all trading updates (buy/sell) but does not store that data.
The script can be run in a virtualenv (via pipenv) or by using the Docker. For local development the virtualenv is recommended. For production deployments the Docker image is recommended.
You need the sqlite3 library installed on your system. On macOS you can use homebrew: brew install sqlite3.
To install the development environment, clone the repo and run:
$ pipenv install --dev
The --dev flag will install development tools such as py.test.
You can build the Docker image with:
$ docker build -t bitfinex .
The Docker image will not install development tools such as py.test.
You can run the script with the following command:
$ pipenv run python bitfinex/main.py --debug
Or if you prefer Docker:
$ docker run --rm bitfinex python bitfinex/main.py --debug
The --debug flag will print detailed information for each request/response, which is very useful for debugging.
$ pipenv run python bitfinex/main.py --debug
2018-04-22 20:48:49,326 INFO Found 105 symbols
2018-04-22 20:48:49,326 DEBUG Found previous db entries. Resuming from latest
2018-04-22 20:48:49,327 INFO 1/105 | btcusd | Processing from 2013-04-20 02:59:00
2018-04-22 20:48:49,328 DEBUG 2013-04-20T02:59:00+00:00 -> 2013-04-20T19:39:00+00:00
2018-04-22 20:48:49,399 DEBUG Starting new HTTPS connection (1): api.bitfinex.com
2018-04-22 20:48:49,624 DEBUG https://api.bitfinex.com:443 "GET /v2/candles/trade:1m:tBTCUSD/hist?start=1366426740000&end=1366486740000&limit=1000 HTTP/1.1" 200 None
2018-04-22 20:48:49,633 DEBUG Fetched 288 candles
The script will import OHLC data for all symbols defined in the symbols.json file. If the import is started for the first time, the date defined in symbols_trading_start_days.json will be used as a starting point. Otherwise the database is queried for the last date and the import will continue from there on.
The data is saved in a file called bitfinex.sqlite3 in the same directory as the script. You can change it by passing it as an argument:
$ pipenv run python bitfinex/main.py /path/to/my/db.sqlite3
The data is stored in a Sqlite3 database and has a "candles" table with the following structure:
symbol (the trading symbol e.g. btcusd)
time (time in unix timestamp with milliseconds added)
open
close
high
low
volume
Example output:
$ sqlite3 bitfinex.sqlite3
sqlite> select * from candles;
btcusd|1366366980000|122.7|122.4|122.7|122.4|1
btcusd|1366366920000|123.2|122.7|123.2|122.7|82.9
btcusd|1366366800000|123.7|123.7|123.7|123.7|1.1
The data can easily be exported to CSV:
$ sqlite3 bitfinex.sqlite3
sqlite> .headers on
sqlite> .mode csv
sqlite> .output data.csv
sqlite> SELECT * FROM candles;
sqlite> .quit
The data will be in a file called data.csv.
You need to have development dependencies installed (pipenv install --dev).
The tests are stored in the tests directory and can be run with py.test:
$ pipenv run py.test tests/