I'm developing an AI personality system using llama.cpp, and I'm doing everything locally. The orchestration is entirely handled in Bash, organized modularly by profiles such as prompt.txt, config, and launcher.sh. I'm curious if anyone has experience with using Bash to structure prompts or manage contextual memory effectively. I've been implementing Levenshtein distance for string comparison and controlling token usage, and I'd greatly appreciate any technical feedback on my script architecture.
2 Answers
I haven't used llama.cpp specifically, but I think using Bash for your project is interesting! Bash can be pretty handy for orchestrating scripts and keeping everything modular. As for managing contextual memory, you might also want to look into some log management to keep track of token usage. It's great that you're experimenting with Levenshtein distance too! Let us know how it goes!
Google Translate says: "I'm working on an AI personality system using llama.cpp completely locally. All the orchestration is done in Bash, with modularity by profiles (prompt.txt, config, launcher.sh, etc.). Has anyone else worked with Bash to structure prompts or manage contextual memory? I'm using Levenshtein distance + token control, and I'd love to get technical feedback on the script architecture." Honestly, it’s better to post in English for better responses."
Totally agree! Just trying to help out here.