All tests run on an 8-year-old MacBook Air.
Android Studio's logcat is fine. But opening a full IDE just to read logs and debug an error felt like overkill.
I built a lightweight logcat viewer in Rust + Tauri. Then I added a Gemini button next to every error line. One click — AI diagnosis, no copy-pasting, no context switching.
Here's how the AI integration works.
The idea: click the error, get the diagnosis
Every error line in the viewer has a small button. Click it, and the app:
- Pulls that error line plus the surrounding 50–100 lines from a ring buffer
- Strips any PII (IP addresses, emails) before sending
- Sends the context to Gemini with a carefully crafted prompt
- Displays the diagnosis in an overlay panel
No manual copy-pasting. No switching to a browser. The diagnosis appears where the error is.
The ring buffer design
Logcat streams thousands of lines per minute. You can't keep all of them in memory.
The Rust backend maintains a ring buffer of the last 2,000 lines. When the user clicks diagnose, slicing out the relevant context is instant:
pub struct LogRingBuffer {
lines: VecDeque,
capacity: usize,
}
impl LogRingBuffer {
pub fn push(&mut self, line: LogLine) {
if self.lines.len() >= self.capacity {
self.lines.pop_front();
}
self.lines.push_back(line);
}
pub fn context_around(&self, target_idx: usize, window: usize) -> Vec<&LogLine> {
let start = target_idx.saturating_sub(window);
let end = (target_idx + window).min(self.lines.len());
self.lines.range(start..end).collect()
}
}
2,000 lines, instant slice, minimal memory.
The prompt
let system = "You are an Android development specialist. \
Analyze the following logcat output and explain \
the root cause and fix concisely.";
let context_text = context_lines
.iter()
.map(|l| l.raw.as_str())
.collect::>()
.join("\n");
let prompt = format!("{}\n\nLogcat:\n{}", system, context_text);
Simple. The context does the heavy lifting — Gemini doesn't need elaborate prompting when the relevant lines are right there.
Result
Open the viewer, run the app, see an error → click the button → diagnosis in 2–3 seconds. No IDE required.
HiyokoLogcat is free and open source → github.com/hiyoyok/HiyokoLogcat
X → @hiyoyok
This article was originally published by DEV Community and written by hiyoyo.
Read original article on DEV Community