Releases: IXtimes/quizzy
Improvements Release v2.0.1
My follow up release to the first version released last month that makes a few bug fixes and performance/stability improvements as I found needed during my studies. These include:
- Quiz building progress now shows when creating a quiz
- Generating quizzes using more advanced models should not max out on tokens anymore at the cost of much slower generation. This should not impact the performance of the cheapest model, and you should be using that model for most purposes anyways
- Misc. bugs relating to the different views of the app, some preventing features of the app from working properly
- Replace essay difficulty options with the option to grade using the set guidelines or not, since this option seems to make more of a difference in the grading than the difficulty options :p.
This build does not add any "new" features, and I still have some that I want to add, but they'll probs be added next time I need to study for finals as I have no content to really test the features against. ¯_(ツ)_/¯
If I make another update to this build (I am getting fed up with using Tkinter), expect to see new features around Mayish. Otherwise any updates that I push will just be bug fixes that I come across or are made aware of.
Major Release v2.0.0
Major Release v2.0.0
This version of Quizzy, kinda cooked up in too short of a time, adds most of the remaining features that I sought to add to the project upon its initial release back in May.
- Added term-definition and essay type questions to fully address all the possible forms questions can take.
- Term-definition based questions support AI assisted generation, including term to definition matching.
- Essay questions feature custom AI grading schemes to more critically assess your understanding of the content as you become more familiar with it.
- Completely rewrote the backend of the project, allowing for a more "structured" project so new features should not be too insane to add.
- Parallelized many of the generative aspects of the app, substantially increasing its performance in question generate and essay grading.
- Updated model selection to include all of the recent models available through the OpenAI API, useable at varying costs.
- Costs to generate questions likely went up with this version, as now some aspects of quiz generation get they're own prompts or may pass through several AI calls, though this is done to substantially increase the quality of the AI output.
- Encryption of context segments to keep them private
- New context file reading feature that supports parsing the text from ,pdfs, .pptxs, .docxs, and .txt and appends it to the current context segment to make generating it substantially easier.
- And more!
Despite this being my fully actuated release, I still have some features that I want to add to the project in the future, but that might be saved for a spring break venture :p. If anything, expect to see the following if not soon, at some point:
- Compiled releases for MacOS and Linux
- Permutation support (placing blanks in questions and having only those be filled with AI assistance, including random number generation)
- Image library support for custom questions to be based on supplied diagram segments
- Quick import options to quickly generate questions using just raw text.
- Plentiful amounts of bug fixes where need be because I know they're definitely still out there.
That being said, I'll likely release a hotfix release come the end of the year to address any bugs that I find while using the app to study for finals, this mainly includes functionalities where ChatGPT may timeout and freeze the application. However, like I said above, I'm sure there is plenty of stuff I may of missed seeing as this release kinda happened due to time constraints.
Enjoy :).
Full Changelog: v.1.0.2.1...v2.0.0
Release v.1.0.2.1
Minor bug fixes with offline mode condition checking flagging in cases where the user was in online mode (using an AI API key)
Release v.1.0.2
Fixed bugs relating to being able to initiate quiz building with sets that had fewer than 3 questions. This would deadlock the program which is very bad.
Otherwise, I also fixed any bugs that could lead to creating quizzes of less than 3 questions, even if the slider denoting the selection may render to say 0 questions.
For right now, this is the version of Quizzy that I am sticking with for a few months. The next features that I may add will relate to adding matching and free-response question functionality. But for now, I think the app has enough standalone features to be useful for studying subjects.
Signed Windows Release v1.0.1
Changes to this version:
- Executable is now code signed
- Added .spec and version.text file to the source for building
- Updated to use the new 'gpt-4o' model for the GPT 4.0 option in settings as it is more cost efficient and overall better.
Please note any issues that I may have missed in the issues tab, as I had to rewrite some resource imports because my old .spec file that I used with PyInstaller got overridden and I had to reconfigure it.
Initial Windows Release v.1.0.0
Inital release of the Quizzy client. This is my first proper application client so its likely to be laggy and unoptimized in some aspects, but overall this build works fairly well with deck sizes of around 150-200 questions at most. Possibly more.
Please note that any AI functionality built into the application is not mandatory, but does require an OpenAI API key in order to prompt for new questions or missing fields. You can acquire one here for yourself: https://platform.openai.com/api-keys
(also its not code signed or anything yet so it might flag antivirus, if you don't trust the software feel free to look through the source yourself or make your own build)