# Quick start OpenAI API example using Python * url: http://datachild.net/machinelearning/openai-api-python-quickstart * category: machinelearning * published: 2023-05-09 * tags: python, openai * description: How to start using OpenAI API with Python. A simple example of a Python script that generates data based on the OpenAI language model. OpenAI has a simple and friendly API, let's take a quick look at how to start using it with Python. ## Installing libs The only thing we need to install is [openai](https://github.com/openai/openai-python) Python module: ``` pip install openai ``` * `pip install` - installs Python package, * `openai` - module to work with OpenAI. ## Getting API key The second thing we need to do is to get the API key for our app from [this page](https://platform.openai.com/account/api-keys):  ## Testing API Now we can use API to get answers from OpenAI using prompts. The only thing we need to define is the [model](https://platform.openai.com/docs/models) we're going to use. We've picked `text-davinci-003` for our example: ``` import openai def ask(prompt): openai.api_key = '<YOUR-API_KEY>' return openai.Completion.create( model="text-davinci-003", prompt=prompt, )['choices'][0]['text'].strip() print(ask('pi number value, 25 decimals')) ``` ```output 3.141592653589793238462643383279 ``` * `import openai` - load module to work with OpenAI API, * `def ask(prompt)` - we create a custom function to ask questions and return answers from AI, * `openai.api_key` - set our API key to be able to send API requests, * `openai.Completion.create` - create a completion object to get answers for our prompts, * `model=` - choose the relevant model from [available ones](https://platform.openai.com/docs/models/gpt-3-5), * `prompt=` - our prompt (we use function argument here), * `['choices'][0]['text']` - API response will go here, * `.strip()` - remove all whitespaces before and after the text we got from API. This will be it. It's an extremely simple example, but that is what makes OpenAI API so cool - it's that easy. Now let's take a look at a more advanced example we can use in practice. ## Practical example - generating test data One of the powerful OpenAI features is that we can ask to use a certain format for output. For example, we can ask API to generate various user data and format it suitably (e.g., CSV), so our script can understand it. Let's create a function to generate a given number of user-related data records. We'll add the following code to our previous example: ``` from io import StringIO import csv def gen_test(n = 5, fields = ['phone', 'first name', 'last name', 'address', 'email']): prompt = 'generate a CSV list of ' + str(n) + ' random records including: ' + ', '.join(fields); csv_data = ask(prompt) f = StringIO(csv_data) reader = csv.reader(f, delimiter=',') return [row for row in reader] print(gen_test()) ``` ```output [['613-456-2264', 'John', 'Smith', '12 Main Street, Ottawa, ON', ' johnsm@example.com'], ['416-213-7648', 'Bob', 'Taylor', '321 Anywhere Street, Toronto, ON', ' btayl@example.com'], ['905-522-9637', 'Jane', 'Davis', '14 Maple Street, Hamilton, ON', ' jdavis@example.com'], ['250-546-6315', 'Rick', 'Jones', '5 Central Avenue, Vancouver, BC', ' rjones@example.com'], ['514-867-0912', 'Sarah', 'Millar', '7 West Street, Montreal, QC', ' smillar@example.com']] ``` * `import csv` - package to work with CSV, * `gen_test(` - the name of the function that returns a list of generated records, * `n =` - number of records to generate, * `fields =` - list of fields to generate for each record, * `prompt =` - prepare a prompt for the API, * `ask(prompt)` - we use a previously created function to retrieve data from AI API, * `csv.reader` - parse `csv_data` (returned text from API) as CSV, * `row for row in reader` - creates a list of parsed records. Now we can try generating data with custom arguments: ``` print(gen_test(10, ['email', 'username', 'password'])) ``` ```output [['email', 'username', 'password'], ['sales@hotmail.com', 'salesperson', 'xhA6sdk4'], ['media@gmail.com', 'mediaperson', '$q3YUjXa'], ['outreach@outlook.com', 'outreach', 'CgE5i#Nj'], ['trade@yahoo.com', 'trader', 'X5nd*bz0'], ['accounting@gmail.com', 'accountant', 'wF0$Dol7'], ['hr@outlook.com', 'hrmanager', 'gF2yD3*h'], ['marketing@hotmail.com', 'marketer', 'K5hM1#r9'], ['consulting@yahoo.com', 'consultant', 'd!$92Hp1'], ['businessdev@gmail.com', 'bdmanager', '2t#Gopn0'], ['engineering@outlook.com', 'engineer', 'TfT%V7au']] ``` The cool thing here is we can put any imaginable fields as we work with a very smart AI: ``` print(gen_test(3, ['First Name', 'Birth Year', 'Credit Card Number', 'Favorite Movie'])) ``` ```output [['Katie', '1986', '4916287068583236', 'Up'], ['Matthew', '1988', '4485709366735621', 'Toy Story'], ['Mia', '1976', '6011938424765310', 'The Godfather']] ``` ## Further reading * [OpenAI Python module repository](https://github.com/openai/openai-python) * [List of OpenAI models that can be used in API](https://platform.openai.com/docs/models) * [Formatting unstructured data using OpenAI API and Python](http://datachild.net/machinelearning/how-to-format-data-using-openai-api) * (@Plan: Extracting entities from unstructured data using OpenAI API) * (@Plan: Searching text documents using vector database and OpenAI API)Save

Upload