A script to detect job rejections from your Gmail inbox.
I've applied to lots of jobs.. easily in thousands by now. Anyway, I'm tired of having my inbox flooded by rejections and wanted to create something that will solve this problem for me.
Gmail has a feature that allows incoming mail to be filtered. I used this feature before starting the project, and realized that there were many limitation to the Gmail query filter: there is a limit to the amount of keywords I can use, and using too many keywords will produce incorrect results.
For example, I might get an email that happens to have the phrases "not considered" and "unfortunately" but it can be anything-- not just related to job search.
This is when I though about AI and how I can utilize it to read my emails and determine if they are job-related rejections.
- A connection to the Google API is established using OAuth credentials.
- The Gmail service is contacted.
- Any new and unread emails are pulled. There is a default limit of the 10 most recent emails. This limit can be changed within the code.
- Emails are parsed using BeautifulSoup to extract all the paragraphs within it.
- A connection is established to OpenAI.
- The emails pulled are fed into a function where GPT 3.5 will determine if an email is a job rejection.
- Emails marked as "rejected" will have a label applied to them.
To run the program, you will need
- a Google and OpenAI account,
- to create a Google Cloud Application and
- generate an OpenAI API token.
Follow the steps in the Google Cloud documentation to create an application.
After creating your application, you'll receive a credentials.json file. You can
review credentials.example.json to see what credentials.json should look like.
Refer to the OpenAI API documentation to get an API token. I have my OpenAI API token loaded into an environment variable.
Install all the dependencies within the requirements.txt file.
Here's a project roadmap to see what's available:
- Google API integration for Gmail authorization
- OpenAI integration for AI usage
- Job rejection status output
- Script may be ran as a CRON job for constant updates
- Indication of which emails are being processed
- Fetched mail and labels are cached locally to reduce wait times