Little news about my JOEF. Yesterday I was taking some time to code (with the help of AI) a Bot that:
1. Gets the job posting records which "status" cell is blank from my job offers spreadsheet.
2. Gets each job posting web page content and parse it to a structured data.
3. Queries a custom llama3.2 model locally via Ollama to evaluate if the job opportunity is a match or not.
It is notable how LinkedIn sends me offers in Spanish, Polish, German, etc. All languages not even mentioned in my profile.
To filter out these ridiculous situations is one of the reasons I begun JOEF project.
The "NO", "YES" and "MAYBE" in the sheet are from this new Bot.
Playwright might not be necessary. Gemini suggested to use it, convinced that it was the only way to circumvent LinkedIn anti-bot gates.
As usual, Gemini (just as like Copilot and the others) made me waste a lot of time, because of this assumption.
The cause was much simpler than that. Somehow, when JOEF main app stores a job posting link in the sheet, the cell actual value is different from the display value (i.e. formatted value). And, in the case of LinkedIn, the actual value is for authenticated agents. The public link, suitable for non-authenticated agents, is the formatted value.
I guess that JOEF fills the cell with the actual value. The display value is then generated by Google Sheet automatically via link analysis.
... Which is an issue that will need further investigation.