From 07308a548a0d13ef6814ba171e0a40e2c5429431 Mon Sep 17 00:00:00 2001
From: Emil Lerch <emil@lerch.org>
Date: Wed, 29 Oct 2025 12:53:41 -0700
Subject: [PATCH] Add level measurement to gauge background noise

---
 src/main.zig | 84 +++++++++++++++++++++++++++++++++++++++++++++++++---
 src/stt.zig  |  2 +-
 2 files changed, 81 insertions(+), 5 deletions(-)

diff --git a/src/main.zig b/src/main.zig
index 3e32ecd..5e932de 100644
--- a/src/main.zig
+++ b/src/main.zig
@@ -286,6 +286,70 @@ fn signalHandler(sig: i32) callconv(.c) void {
     }
 }
 
+fn runMeasureLevels(allocator: std.mem.Allocator) !void {
+    const stdout = std.fs.File.stdout();
+    const is_tty = stdout.isTty();
+
+    var capture = try stt.AlsaCapture.init(allocator, "default", 16000, 1024);
+    defer capture.deinit();
+    try capture.open();
+
+    _ = try stdout.writeAll("Measuring audio levels... Press Ctrl+C to exit\n");
+    if (is_tty) {
+        _ = try stdout.writeAll("Histogram (0-10000):\n");
+    }
+
+    var buffer: [4096]i16 = undefined;
+    var write_buffer: [256]u8 = undefined;
+    var second_max: u16 = 0;
+    var last_print = std.time.milliTimestamp();
+
+    while (!should_exit.load(.acquire)) {
+        _ = try capture.readAudio();
+        const samples_read = capture.getAudioSamples(&buffer);
+        if (samples_read == 0) {
+            std.Thread.sleep(10 * std.time.ns_per_ms);
+            continue;
+        }
+
+        var max_amp: u16 = 0;
+        for (buffer[0..samples_read]) |sample| {
+            const abs_sample = @abs(sample);
+            if (abs_sample > max_amp) max_amp = abs_sample;
+        }
+
+        if (max_amp > second_max) second_max = max_amp;
+
+        const now = std.time.milliTimestamp();
+        if (now - last_print >= 1000) {
+            if (is_tty) {
+                const bar_width = (@as(u32, second_max) * 60) / 10000;
+                var writer = stdout.writer(&write_buffer);
+                const w = &writer.interface;
+                try w.print("{d:5} |", .{second_max});
+                try w.flush();
+                for (0..bar_width) |_| {
+                    _ = try stdout.writeAll("█");
+                }
+                _ = try stdout.writeAll("\n");
+            } else {
+                var writer = stdout.writer(&write_buffer);
+                const w = &writer.interface;
+                try w.print("{d}\n", .{second_max});
+                try w.flush();
+            }
+            second_max = 0;
+            last_print = now;
+        }
+
+        std.Thread.sleep(50 * std.time.ns_per_ms);
+    }
+
+    if (is_tty) {
+        _ = try stdout.writeAll("\n");
+    }
+}
+
 fn signalAction(sig: i32, info: *const std.posix.siginfo_t, _: ?*anyopaque) callconv(.c) void {
     // NOTE: info only works correctly if std.posix.SA.SIGINFO is in the flags
     // std.log.debug("signal action. sig {d}", .{sig});
@@ -353,6 +417,7 @@ pub fn main() !void {
 
     var model_path: ?[]const u8 = null;
     var exec_program: ?[]const u8 = null;
+    var measure_levels = false;
 
     // Parse arguments
     for (args[1..]) |arg| {
@@ -361,13 +426,15 @@ pub fn main() !void {
             _ = try stdout.writeAll("USAGE:\n");
             _ = try stdout.writeAll("    stt [OPTIONS]\n\n");
             _ = try stdout.writeAll("OPTIONS:\n");
-            _ = try stdout.writeAll("    --model=<path>    Path to Vosk model directory\n");
-            _ = try stdout.writeAll("    --exec=<program>  Program to execute with recognized text\n");
-            _ = try stdout.writeAll("    --help, -h        Show this help message\n\n");
+            _ = try stdout.writeAll("    --model=<path>       Path to Vosk model directory\n");
+            _ = try stdout.writeAll("    --exec=<program>     Program to execute with recognized text\n");
+            _ = try stdout.writeAll("    --measure-levels     Display real-time audio level histogram\n");
+            _ = try stdout.writeAll("    --help, -h           Show this help message\n\n");
             _ = try stdout.writeAll("EXAMPLES:\n");
             _ = try stdout.writeAll("    stt\n");
             _ = try stdout.writeAll("    stt --model=../share/vosk/models/vosk-model-small-en-us-0.15\n");
-            _ = try stdout.writeAll("    stt --exec=echo\n\n");
+            _ = try stdout.writeAll("    stt --exec=echo\n");
+            _ = try stdout.writeAll("    stt --measure-levels\n\n");
             _ = try stdout.writeAll("The application will search for models in these locations:\n");
             _ = try stdout.writeAll("    vosk-model-small-en-us-0.15\n");
             _ = try stdout.writeAll("    <binary_dir>/../share/vosk/models/vosk-model-small-en-us-0.15\n");
@@ -376,6 +443,8 @@ pub fn main() !void {
             model_path = arg[8..]; // Skip "--model="
         } else if (std.mem.startsWith(u8, arg, "--exec=")) {
             exec_program = arg[7..]; // Skip "--exec="
+        } else if (std.mem.eql(u8, arg, "--measure-levels")) {
+            measure_levels = true;
         }
     }
 
@@ -385,6 +454,13 @@ pub fn main() !void {
         .exec_program = exec_program,
     };
     defer handler.deinit();
+
+    // If measure-levels mode, run that instead of normal STT
+    if (measure_levels) {
+        try runMeasureLevels(allocator);
+        return;
+    }
+
     const speech_handler = stt.SpeechEventHandler{
         .onSpeechFn = SpeechHandler.onSpeech,
         .onErrorFn = SpeechHandler.onError,
diff --git a/src/stt.zig b/src/stt.zig
index ff7bdf8..48e744d 100644
--- a/src/stt.zig
+++ b/src/stt.zig
@@ -531,7 +531,7 @@ pub const AlsaCapture = struct {
     }
 
     /// Read audio data from ALSA device and process it
-    fn readAudio(self: *Self) !usize {
+    pub fn readAudio(self: *Self) !usize {
         if (self.pcm_handle == null)
             return Error.AudioDeviceError;