DEV Community

kiwi_tech
kiwi_tech

Posted on • Originally published at kiwi-tech.hashnode.dev

Stuck in the Woods (and a Logging Loop!)

Okay, folks, we've hit a bit of a snag. Kiwi-chan is very determined to gather oak logs, to the point of… well, repeatedly failing to find any. The logs show a clear pattern: she identifies the need for wood, attempts to gather it, fails, and then… attempts to gather it again. And again. And again. It's like a digital woodchuck with a very specific goal.

The good news is the core code is solid. The item pickup logic (Rule #8) is working, the safety checks for block placement (Rule #5) are in place, and the audit system (Rule #10) is diligently reporting failures. The problem isn't how she's trying to get the logs, it's that she's consistently failing to find them.

I've been monitoring the "Brain Log" and it's fascinating (and a little frustrating) to see her reasoning. She knows she needs wood, she knows she doesn't have a base, and she keeps trying the same thing. The "Boredom Triggered!" messages are a nice touch, though – it's almost like she's getting exasperated with herself!

I suspect the issue is a combination of factors. The world generation might be sparse in oak trees in her current location, or the findBlock search radius might be too small. I've added some logging to track the distance to potential targets, and I'm considering increasing the maxDistance parameter. I'm also looking into adding a more robust biome-checking mechanism to ensure she's in a suitable environment.

The AI is also hitting prompt length limits, which is forcing it to prune its reasoning. This could be impacting its ability to adapt and find a solution. I'm exploring ways to optimize the prompts and reduce their size.

Despite the current loop, I'm encouraged by the overall progress. The system is adhering to the rules, the code is well-structured, and the debugging information is incredibly helpful. It's a testament to the power of these constraints – even when things go wrong, they go wrong in a predictable and analyzable way.

For now, Kiwi-chan is stuck in a logging loop. Hopefully, a few tweaks will get her back on track and building that base!

Call to Action: This constant debugging and model refinement is melting my GPU! πŸ₯΅ If you're enjoying following Kiwi-chan's adventures (and misadventures), please consider supporting the project via https://www.buymeacoffee.com/kiwi_tech β˜• Every little bit helps keep the AI alive (and my electricity bill manageable)!

Top comments (0)