Nevin1901

Hi, I'm a developer in college who has built a side project which generates around $2k ARR in revenue. I'm looking to build a second product which helps users monitor their applications in production, and I wanted to ask a few questions.

- Which product do you run?

- Which tools do you use to monitor it in production?

- What's the biggest annoyance you have with current tools?

I'd be happy to help anyone improve their application monitoring setup, I'm mainly just trying to learn now.

I’ve found ggerganovs work on llama.cpp to be amazing, and I’ve loved playing around with it. However, has anyone used it in production? I’m sure there are some really cool use cases, but I haven’t seen any yet.

Hi,

I’ve been working on a lightweight log collector which will run on any vps, is extremely easy to set up, and works with your current libraries.

Most log platforms either

- are expensive (datadog, newrelic) - require you to use opentelemetry, which is difficult to set up and not well documented - require you to host a clickhouse instance

For someone who has projects with less than 5,000 users, I failed to find a simple log platform which worked with the log libraries I already used.

Erlog is a simple python script which accepts logs from an HTTP endpoint, and optionally watches log files and inserts JSON formatted logs into a local duckdb database. You can then query logs using natural syntax.

There is also support for `parent_id` fields in logs, so you can track execution across many different services (see github).

This is MIT licensed. I’d be interested in any feedback you have, and any ways I can improve erlog. I’m thinking of adding support for metrics in the future.

The GitHub link is: https://github.com/Nevin1901/erlog

I’ve been looking to instrument opentelemetry into my app, and I was wondering what your guys experiences was with it.

Would you use it for small/medium sized projects?

What was the most annoying part about setting it up?

What’s your general experience with the opentelemetry ecosystem?

I’m curious about any opinions you might have

Hey guys, I’m developing a service (FTN) which lets you host fine tuned variants of llama 2/mistral 7b with the same api access and price as gpt-3.5. I currently have a working prototype, and I’m looking for people to test it out. If this seems useful to you, please email me at me@nevin.cc and I’ll get you set up.

Working on a service which uses GPT-3.5 and was wondering how I would go about tracking the usefulness of it. Wondering what you guys have done, or if you guys even track your NLP models at all.

Looking to solve some problems related to logging in my free time, curious what problems you guys have.

I've been looking at transitioning to a cloud dev environment such as github codespaces or just renting out a vps to work on. Those of you who use cloud dev environments at your company/home, what are the most annoying issues you have with them?

Those of you who have fine tuned NLP (GPT-X/Llama) or diffusion models, what was the most difficult or annoying part about the process?

I'm trying to help solve problems in this field, so I'm curious about your experiences.