Dragonish

I am interested in reading more about what Tezka means. Please do share.

I think i can relate to your goals and am personally focused on similar work in an effort to make my own life a little more bearable. my efforts are more focused on executive function and how to integrate this into my life seamlessly vs llm/conversational ai. i have been playing around with conversational ai, but i currently lack the psychological understanding which is needed to do this right. i look forward to hearing more from you.

my immediate (ok, i have been working on this all day) thoughts

• as other have mentioned, i like quirky. I would want them to show some flaws. idk what exactly, but i think it would be off-putting to be overly clinical or “perfect”
• i would be more comfortable interacting with Tezka in a more private environment such as a matrix room vs a more public comm like this.
• i like the “relational repair” aspect. my own shortcomings here is something that has been made much clearer to me recently. I imagine them asking me if i have reached out to my relations, and give me some personalized advice on how best to approach the person. If the interaction with the person did not go well, then i imagine them helping me through it in a positive way, preparing me to try again next time.

I am using Mistral 7b Instruct for text summary and some light “assistant” type chatting for the last several months. I have been pleased at how accurate it is for my needs, especially given it’s size.

I recall alot of trial and error to find models that were compatible with the version of llama-cpp-python that oobabooga uses (at any given time). GGUF should have made the model format (and therefore model selection) more simple, but i imagine there are still nuances that make it more difficult than it should be to find a working model for a noob.

Best of luck, let us know how it goes

VS Code’s extension system makes it pretty easy to build your own code snippet extension. I use my own private extension to easily “generate” different types of markdown files (ie readme vs a troubleshooting guide) from my personalized snippets.

I agree that data staleness is a limiting factor. Depending on your needs and technical proficiency you could use use their zimit service (limited in the number of links it follows). The zimit tool is oss and on github, so you can run the it yourself to keep the sites you are interested in up to date in your local kiwix

While not a solution right now, I just want to add that the general transit feed spec aims to solve the data interoperability of different transit systems. The transit system keeps a publicly accessible zip file up to date, and then anyone can pull/parse the schedule/prediction data in a consistent way across transit systems. I know in the us adoption is slow, with vendors prefering to build their own walled gardens and transit agencies lacking the vocabulary or skills to advocate for more open data/tools