The link to the PR - https://github.com/langflow-ai/langflow/pull/5261

Profilers help if you want to identify the most time-consuming parts of your codebase, but there isn’t a tool we can use to understand how much faster our *overall software* can get with the right tweaks.

Does anyone know of a solution to this problem? Or will this always be manual?

1. I didn’t think enough.
2. I didn’t think this would be a problem
3. The way the software is used now is quite different from what I originally expected.

When it comes to performance optimization, you only know you’ve got slow code when you find a better version that runs faster.

By aggregating these samples, PyInstrument identifies key “hot spots” where your code spends the most time, presenting the data through very intuitive visualizations like flame graphs.

Shock me - what’s the most expensive workload you’ve run?

Exactly! These non functional requirements are hard to argue for by engineers. That's why I am building codeflash.ai which aims to automate performance optimization so that engineers can focus on shipping features, while AI ensures that it runs fast as well.

I'm eager to hear from other engineers: How are you addressing performance challenges in your organizations? What strategies have you found effective in maintaining software efficiency without sacrificing development speed?

With Codeflash, I'm exploring a more proactive approach to performance engineering. The goal is to build performance consciousness into our development workflows, preventing technical debt before it becomes a systemic problem.

The pattern is frustratingly predictable:

1. Ship features quickly
2. Performance gradually degrades
3. Users complain or infrastructure costs spike
4. Briefly optimize performance
5. Repeat the cycle

Performance ideally should be a first-class citizen in our dev process, not an afterthought.