Devlog

2026-03-09

What I worked on:

• Implemented the storage selection system for the PC builder with Entry / Mid / High tiers.
• Connected storage logic to the purpose rules table so storage tier can shift depending on workload.
• Added fallback logic so storage can downgrade tiers if the preferred option does not fit the budget.
• Implemented the minimum build fallback system that generates the cheapest valid PC build when the normal build cannot fit the user’s budget.
• Created calculateMinimumRequiredBudget() to find the cheapest valid combination of GPU, CPU, PSU, motherboard, RAM, and storage.
• Implemented recommended normal budget detection using calculateRecommendedNormalBudget() which finds the lowest budget where the normal builder succeeds.
• Added upgrade suggestion messages for users when they receive a minimum fallback build.

Problems / Debugging:

• Encountered a maximum call stack size exceeded error caused by recursive calls between the budget recommendation system and the build engine.
• Fixed the recursion issue by adding a control flag so the budget recommendation function can test builds without triggering additional recommendation calculations.

What I learned:

• How to design a fallback system when constraints prevent the normal algorithm from producing a result.
• How recursive functions can unintentionally cause infinite loops when two systems call each other.
• How to structure a backend API so the frontend can clearly understand build modes (normal vs minimum-fallback).

Next:

• Complete the Framer UI to display build results.
• Show part cards for GPU, CPU, PSU, motherboard, RAM, and storage.
• Display upgrade suggestions when the builder returns a minimum fallback build.

i added a PSU, RAM, and downgrade system and and a limit to not downgrade too much and eventually lower the performance

I first prototyped the logic using automation tools and a database to figure out how a real PC builder allocates budget between CPU, GPU, and other components. After testing, I started migrating the system into a real backend.

So far I:
• Designed the frontend UI in Framer
• Built a rule-based configuration engine
• Connected an AI model that extracts budget, purpose, and performance tier from user text
• Moved the database from Airtable to a local JSON database
• Rewrote the automation logic into a Cloudflare Worker API
• Implemented GPU and CPU selection with fallback safety (downgrades if the budget is too low)

Right now the system can take a request and generate a realistic PC build instead of random recommendations.

Next I’m working on adding the rest of the components (RAM, motherboard, PSU, storage) and improving the result page. Later I want to add a 3D preview so users can visualize how the PC will look.

I attached a demo video showing the concept and early prototype. Feedback is very welcome!