Skip to content

DEVLOG.md

Latest commit

 

History

History
91 lines (70 loc) · 2.79 KB

DEVLOG.md

File metadata and controls

91 lines (70 loc) · 2.79 KB

Note: All time is in UTC

2026/02/16

  • 17:01

    • The project has been created and the base documentation files and structure have been created.

  • 15:18

    • The README file has been created and the Local LMs have been installed and tested for functionality.

  • 21:20

    • Created the get_state.py file to obtain the state of the webpage via the URL.

  • 21:50

    • Added a robots.txt checker and a scroll function to avoid "lazy-loading" to get the most out of the data.

  • 04:52

    • Created the processor.py file for the Visual Processing module.

  • 06:53

    • The processor.py module now processes possible candidates for buttons in the image provided.

  • 20:03

    • Created the skeleton of the classifier.py module.
    • Updated the comments and documentation.

  • 20:58

    • Finished the base of the cassifier.py module and now it can function.
    • Now creating a scout.py file to orchestrate the pipelines.

  • 21:34

    • It works as intended and can now read the text on the screen.

2026/02/18

  • 17:04

    • Staring an AI code audit and found the following on scout.py:
      • Keep browser open for agentic actions.
      • Rename scrape to observe for proper naming convention.
      • Need error handling on get_web_state() to avoid crashes.
      • Lacks memory, a current_state variable could help.
      • logging is better than print in these circumstances.

  • 17:44

    • Refactored scout.py and now moving on to get_state.py.
    • get_state.py needs the following changes:
      • The browser needs a seperate class to persist.
      • Add a network idle to ensure all is loaded.
      • Returning the page as well as the screenshot is better.

  • 18:18

    • Refactored the get_state.py module now moving onto processing.py.
    • The following needs to be done for processing.py:
      • Keep the overlapping box that makes the most sense.
      • Safety check to avoid the ValueError that would crash the program.
      • Read image once pass it around for efficiency.

  • 18:50

    • Refactored the processing.py module and now moving on to the classifier.py after a short break.

  • 19:56

    • The classifier.py module needs the following refactors:
      • Regex cleaning to strip common OCR hallucinations.
      • Log instead of printing.
      • --psm 6 might be bettter than --psm 7 for the config.

  • 20:17

    • Refactored the classifier.py module.

  • 21:56

    • About to step away soon, need to finish the artifacts that appear.

2026/02/19

  • 21:44
    • Picking away at finding each element with many filters.

2026/02/23

  • 21:11
    • Created the ARCHITECTURE.md file so I can stop playing with values and try a more calculated aproach... Which I should have done to begin with.

2026/02/24

  • 05:22 A rewrite based on the architecture has been finished and the elements are properly found.

TO-DO