Let's create some sample data, something GPT is great at. ggpt prompt will take our input and return GPT's output.
$ ggpt prompt "Output a sample csv file of new cars for sale. Include car model, cost, and mpg."
model,cost,mpg
Toyota Camry,25000,32
Honda Civic,22000,36
Ford Fusion,26000,28If the output is sufficient, ggpt last will return it— no need to query GPT again. Lets save it to for_sale.csv.
$ ggpt last > for_sale.csvTaking advantage of command substitution, GPT can easily parse over this new data. Let's get GPT to write us some Python to analyze our data.
$ ggpt prompt "Output an example Python file that reads in a file named for_sale.csv with data $(cat for_sale.csv), and prints the cost per mpg for each car model."
import csv
with open('for_sale.csv', 'r') as file:
reader = csv.reader(file)
next(reader) #Skip header row
for row in reader:
cost_per_mpg = float(row[1]) / float(row[2])
print(row[0] + ": " + str(cost_per_mpg))
$ ggpt last > cars.py && python3 cars.py
Toyota Camry: 781.25
Honda Civic: 611.1111111111111
Ford Fusion: 928.5714285714286Another possible usage of GPT is regex help (we all need it).
$ ggpt prompt "Output a regex that finds all december dates from 2011 to 2019. Use format month/day/year. Include no explanation."
\b12\/([1-9]|[1-2][0-9]|3[0-1])\/(201[1-9])\bCool, that's certainly a regex. We know GPT can (and will) make stuff up. Lets save that regex to an environment variable, and test it.
$ export REGEX=$(ggpt last)
$ ggpt prompt "Output a list of 20 dates, one on each line. Use format month/day/year. Include dates from the last ten years." | tee dates.csv
Assuming the current year is 2021, here is a list of 20 dates in the format month/day/year, including dates from the last ten years:
1. 03/14/2011
2. 11/11/2012
3. 05/16/2013
4. 08/08/2014
5. 01/27/2015
6. 06/12/2016
7. 12/31/2017
8. 09/22/2018
9. 04/09/2019
10. 11/03/2020
11. 01/10/2021
12. 04/22/2021
13. 07/04/2021
14. 09/15/2021
15. 11/25/2021
16. 01/01/2022
17. 03/17/2022
18. 06/19/2022
19. 08/14/2022
20. 12/01/2022
$ grep -E "$REGEX" dates.txt
7. 12/31/2017Passes the eye test, pretty neat. What's it doing though?
ggpt prompt "can you explain what regex $REGEX is doing?"
The regex \b12\/([1-9]|[1-2][0-9]|3[0-1])\/(201[1-9])\b is looking for a date in the format "dd/mm/yyyy" where the month is specified as "12".
Here is a breakdown of the different parts of the regex:
\b - Matches the start or end of a word boundary
12\/ - Matches the string "12/" exactly
([1-9]|[1-2][0-9]|3[0-1]) - Matches a string of digits representing the day of the month from 1 to 31
\/ - Matches a forward slash character
(201[1-9]) - Matches a string of digits representing the year from 2011 to 2019
\b - Matches the end of a word boundary
Overall, the regex searches for any string of text that meets the aforementioned conditions, and returns that string as a match.- Make sure golang is installed using
go version. If not, install go.
$ go version
go version go1.20.0 linux/amd64- Install ggpt using
go install:
$ go install github.com/islewis/ggpt@latestThat's it! To confirm ggpt is downloaded, run ggpt --help:
$ ggpt --help
ggpt is a CLI tool to interact with OpenAI's GPT language model. ggpt wraps OpenAI's completion feature, via their API, outputting the result directly in the terminal.
Usage:
ggpt [command]
Available Commands:
completion Generate the autocompletion script for the specified shell
configure Configure your OpenAI API key
help Help about any command
last Returns the output of the previous query.
prompt Call GPT autocomplete with the given string as a prompt
Flags:
-h, --help help for ggpt
-t, --toggle Help message for toggle
Use "ggpt [command] --help" for more information about a command.In order to access GPT, you need an OpenAI account, and an API key. Get one here if you don't already have one.
$ ggpt configure
OpenAI API Key:
Key set