I was building FIE "an open source LLM monitoring system" and I was so proud of myself. The architecture was clean, the endpoints were working, everything looked good.
Then I tried calling my own API from a Jupyter notebook.
"Clean 404. Every single time".
I spent the next 2-3 hours convinced something was seriously broken. I checked the server logs. I rewrote the request function. I restarted the server. I checked the port number. I even started questioning whether FastAPI was the right choice.
At one point I genuinely thought maybe I am not cut out for this.
The actual problem?
My notebook was calling /monitor
My router was mounted at /api/v1/monitor
7 characters. That was it. A prefix I had written myself, that I knew existed, that I somehow never thought to check because I was so sure the bug had to be something serious.
The more complex your project gets, the more you assume the bug must be complex too. Sometimes it is just /api/v1.
Now I really want to know from you all — what is the silliest bug that ate the most of your time?
Or am I the only one who goes through these things?
Top comments (0)