[v4.x-v5] Documentation & Roadmap #17
Description
Due to how fast the project moved from PoC to v4.2, most of the current documentation is not aligned with the code/prompts and the decisions made. Most of the documentation needs refactoring in crucial parts.
For example:
- PDF Metadata optimization was a planned feature that I couldn't get around to implement, but is mentioned as a feature in the docs. By making it look like the PDFs came from Adobe Acrobat, these documents are no longer prone to forensic anti-AI detection
- The version evolution in the README.md is misrepresented
- The existence of Jobbernaut Extract to go along with Tailor is not mentioned anywhere in the docs.
- Retro-fitting the system to your own resume, template, hallucination risk tolerance nature, and prompts while doable is a non-trivial problem currently. In a way while it increases intended friction, the amount of hardcoding also reflects badly on me as an engineer.
Most of the hard work was done in v1, and v2 was the stable highly creative usable engine, v3 was the "works all the time for sure" edition, but now slowly the project is taking a different path of mass scalability (k8s clusters) without cost and quality compromises, so the documentation has to reflect that because eventually it has to migrate away from APIs for the most part and train/fine-tune specialized models for each task.
v5 is planned to be the realization of this vision. I have built an extremely good factory. Now, I want to build a giga-factory. This is where you'll be seeing Redis caching, specialized small models, extreme scalability, high throughput, and ensured recovery. This release could take 5-6 months as I'm getting busier in life and have to go through a lot of engineering blogs and system design patterns to scale at that level. I don't know any of that yet, but I'm confident I'll learn just like I did currently. Most importantly, the goal is to unlock "hostability"
The code is also getting clunkier, and little hard to debug because of the rapid pace of development incurring technical debt (which was planned, it isn't groundbreakingly bad just annoying) and will be fully modularized and refactored in v4.x
Would greatly appreciate assistance in keeping the technical barrier and cost of entry high while keeping it ultra smooth post-entry.