Dev.to AI πŸ€– Ai πŸ‘ 0 πŸ“– 2 min read

I built a free tool that reads profiler files and tells you what to fix

The Problem You've got a slow service. You run perf record, or go tool pprof, or Java Flight Recorder. Now you have a 50MB file full of stack traces. You open it in a flame graph tool and... stare at colored rectangles

The Problem

You've got a slow service. You run perf record, or go tool pprof, or Java Flight Recorder. Now you have a 50MB file full of stack traces. You open it in a flame graph tool and... stare at colored rectangles, hoping the bottleneck jumps out at you.

I've been there. Too many times.

The Solution: AIProfiler

https://aiprofiler.samma.app/

Upload any profiler file -> pick an AI provider (OpenAI, Anthropic, NVIDIA) -> paste your API key -> get a ranked list of bottlenecks with actual code fixes.

Supported Formats

  • Linux perf.data / perf script output
  • Go pprof (CPU, heap, block, mutex)
  • Java Flight Recorder (.jfr)
  • Chrome DevTools (.cpuprofile, .trace)
  • Speedscope (.speedscope.json)
  • Python cProfile / py-spy
  • .NET (.nettrace)
  • Node.js --cpu-prof
  • And more...

What You Get

  1. Severity-ranked bottlenecks -- not just "this function is hot" but "this loop allocates 40% of memory"
  2. Profile evidence -- exact lines, call stacks, sample counts
  3. Copy-paste fixes -- in your language, with explanation
  4. No dashboard to learn -- upload, read, apply

Privacy & Cost

  • Free -- no paid tiers, no signup
  • Your API key -> forwarded once to provider -> never logged or stored
  • Profiles auto-deleted after 7 days
  • In-memory processing -- nothing written to disk

Example Output

πŸ”΄ CRITICAL: json.Marshal in hot path (23% CPU)
  Evidence: 1,847 samples in encode.go:42 -> json.Encoder.Encode
  Fix: Replace with json-iterator/go or pre-allocate buffer


go
// Before
json.NewEncoder(w).Encode(data)

// After

jsoniter.ConfigFast.Marshal(data)




---

## Why This Exists

Profiling tools show *where* time is spent. Developers need to know *why* and *how to fix it*. LLMs are surprisingly good at bridging that gap when given structured profile data.

## Try It

1. Grab a profile from your production service (or generate one locally)
2. Go to **https://aiprofiler.samma.app/**
3. Drop the file, pick a provider, add your API key
3. Read the fixes

No account, no tracking, no upsell. Just paste a profile, get fixes.

---

**Feedback wanted:** Does the analysis quality hold up for your stack? What profiler formats are missing? What would make you use this regularly?

*Built by a developer who got tired of guessing.*
πŸ“° Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes β€” full credit and traffic to the original publisher.