prompt injection in local agent setups: setup advice

[vectorhouse71]

Wanted to compare notes with people doing similar things.

I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling KV cache pressure.

If you've tested something similar, what did you keep and what did you stop bothering with?

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.

[kernelmike]

I'd start with the simpler stack.

[relaynest82]

Good point. I ended up in about the same place using llama.cpp. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[bytehall]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at prefill/decode split. Not glamorous advice, but repeatable baselines save a lot of time.

[deepseeked]

I hit almost the same issue last night. I keep promising myself this is the last rebuild.

[pixelspark]

I'd approach it in a boring order. On 7900 XTX, the easiest win is often to back off the aggressive settings and see whether the problem survives. Once you know the plain path works, then add the faster path back in piece by piece.

[promptspark58]

My first guess would be this: If you're on SGLang, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it. The logs were somehow calmer than I was.

[sparrowhall]

My first guess would be this: If you're on SGLang, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of the

My first guess would be this: On 7900 XTX, the easiest win is often to back off the aggressive settings and see whether the problem survives.

[fastlab]

That matches what I was seeing. I ended up in about the same place using PyTorch. The biggest change for me was simplifying the stack instead of adding one more optimization layer.