Add sine-tone filler with UDP/TCP gap detection; bump to v0.1.2.

- New --filler-mode {silence|tone} and --filler-tone-hz (default 1000) flags.
  Default stays silence so existing setups are unaffected.
- UDP and TCP recv loops now use SO_RCVTIMEO at chunk-period cadence; after
  --filler-after-ms (default 500 ms) of no input, the reader emits filler
  chunks at the audio sample rate. Resumes immediately on real packets.
- FillerGenerator carries phase across chunks so the tone has no clicks at
  chunk boundaries; tests cover silence/tone modes, amplitude (-20 dBFS),
  frequency (zero-crossing count), and phase continuity.
- README documents the new flags, the three trigger conditions, and a Gqrx
  end-to-end example.
- Refined Gqrx UDP example: drop the ffmpeg passthrough (read UDP directly),
  add HTTP Basic auth + --silence-dither + --keep-alive, plus an
  --auth-password-env follow-up and an iPhone Safari credential note.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Author: dsward2

README.md

@@ -326,6 +326,20 @@ Options:
                             treating the stream as dead.
   --silence-dither-ms <n>   Milliseconds of pure-silence before dither kicks
                             in (default: 500).
+  --filler-mode <mode>      What to broadcast during the keep-alive
+                            silence-fill window after stdin EOF.
+                            "silence" (default) emits digital zero
+                            (optionally TPDF-dithered via --silence-dither).
+                            "tone" emits a continuous sine wave at
+                            --filler-tone-hz so listeners hear an audible
+                            "dead-air" placeholder. Only effective with
+                            --keep-alive.
+  --filler-tone-hz <hz>     Frequency of the sine-tone filler in Hz
+                            (default: 1000). Ignored unless --filler-mode tone.
+  --filler-after-ms <n>     Milliseconds of consecutive UDP/TCP input absence
+                            before the filler kicks in (default: 500). Brief
+                            network jitter passes through unaltered; longer
+                            gaps switch to filler.
   -V, --verbose             Verbose logging
   -v, --version             Print version string and exit
   -h, --help                Show this help
@@ -381,13 +395,42 @@ ffmpeg -f avfoundation -i "VB-Cable" \
 
 ### Gqrx UDP audio feed
 
+**Configure Gqrx first.** In Gqrx, click the **UDP** button in the *Audio*
+section. In *Audio Options → Network*, set **UDP host** to `localhost`,
+**UDP port** to `7355`, and tick the **Stereo** checkbox. (Gqrx will start
+streaming as soon as LiveAudioServer is listening on that port.)
+
+Then start LiveAudioServer:
+
 ```bash
-# Gqrx commonly sends 16-bit stereo PCM to UDP port 7355.
-# This example assumes Gqrx UDP output is configured for 2-channel stereo.
-ffmpeg -f s16le -ar 48000 -ac 2 -i "udp://localhost:7355?listen" \
-       -f s16le -ar 48000 -ac 2 - \
-  | .build/release/LiveAudioServer --rate 48000 --channels 2
-```
+# Gqrx sends 48 kHz 16-bit stereo PCM to UDP port 7355 by default — read it
+# directly without any external resampler. Basic-auth gates the HTTP routes,
+# --silence-dither keeps downstream tools alive between transmissions, and
+# --keep-alive holds the stream open across gaps in the upstream feed.
+.build/release/LiveAudioServer \
+    --udp-input-port 7355 \
+    --auth-user alice \
+    --auth-password s3cret \
+    --silence-dither \
+    --keep-alive
+```
+
+A few notes about this example:
+
+- **Don't put the password on the command line in production** — it shows up
+  in `ps` output and shell history. Use `--auth-password-env` instead:
+  ```bash
+  export LIVEAUDIO_AUTH_PW=s3cret
+  .build/release/LiveAudioServer --udp-input-port 7355 \
+      --auth-user alice --auth-password-env LIVEAUDIO_AUTH_PW \
+      --silence-dither --keep-alive
+  ```
+- Basic auth gates **every** HTTP route, including the status page at `/`.
+  The first time you open `http://<mac>.local:8080/` on iPhone Safari you'll
+  see a credential dialog — enter `alice` / `s3cret` (Safari can remember
+  them in the keychain).
+- For anything beyond loopback, pair Basic auth with `--tls-port` so
+  credentials don't ride in plaintext on the LAN (see "Native HTTPS" below).
 
 ### RTL-SDR via `rtl_fm` (mono)
 
@@ -746,7 +789,7 @@ paths and version. `path=/` is the conventional Safari Bonjour-bookmark key;
 `status=/` is the same path under an explicit name for non-Safari clients:
 
 ```
-ver=0.1.1
+ver=0.1.2
 path=/
 status=/
 mp3=/stream.mp3
@@ -759,7 +802,7 @@ details, so a LiveAudioServer-aware client can enumerate all streams in a
 single Bonjour lookup without hitting `/status.json`:
 
 ```
-ver=0.1.1
+ver=0.1.2
 path=/
 status=/
 rate=48000
@@ -823,6 +866,9 @@ Example `server.json`:
   "reopenFIFO": true,
   "silenceDither": false,
   "silenceDitherMs": 500,
+  "fillerMode": "silence",
+  "fillerToneHz": 1000,
+  "fillerAfterMs": 500,
   "verbose": false,
   "authUser": "alice",
   "authPassword": "s3cret",
@@ -860,7 +906,7 @@ This is what `ffmpeg -f s16le` produces, which is the standard raw PCM format.
 
 ### Stream robustness
 
-Three options help the stream survive upstream hiccups:
+Four options help the stream survive upstream hiccups:
 
 - **`--keep-alive`** — after stdin EOF, hold the HTTP outputs open and feed
   the encoders silence (paced at sample-rate) so listeners aren't disconnected.
@@ -873,6 +919,27 @@ Three options help the stream survive upstream hiccups:
   the broadcaster substitutes inaudible TPDF dither (±1 LSB, ≈-90 dBFS), so
   downstream tools that flag pure-silence don't treat the stream as dead.
   Applies equally to stdin, UDP, and TCP input.
+- **`--filler-mode tone`** — replaces the silence emitted during input gaps
+  with an audible sine wave at `--filler-tone-hz` (default 1000 Hz, -20 dBFS).
+  Useful when listeners need a positive signal that the stream is up but the
+  producer is between sources, rather than just hearing dead air. The phase
+  accumulator carries across chunks so the tone has no audible clicks at
+  chunk boundaries.
+
+  Triggers in three scenarios:
+  - **stdin EOF + `--keep-alive`** (legacy behavior).
+  - **UDP input idle**: no packets for `--filler-after-ms` (default 500 ms).
+  - **TCP input idle**: connected client stops sending for `--filler-after-ms`.
+
+  ```bash
+  # Gqrx → LiveAudioServer; if you mute Gqrx's UDP output, listeners hear a
+  # 1 kHz tone after 500 ms instead of dead air, until Gqrx resumes.
+  .build/release/LiveAudioServer \
+      --udp-input-port 7355 \
+      --filler-mode tone \
+      --filler-tone-hz 1000 \
+      --keep-alive
+  ```
 
 ---
 

Sources/LiveAudioServer/Config.swift

@@ -72,6 +72,21 @@ struct ServerConfig {
     /// Number of consecutive all-zero samples (per channel, summed) that must
     /// elapse before dither kicks in. Default ~500 ms at 48 kHz stereo.
     var silenceDitherThresholdSamples: Int = 48_000
+    /// What the reader broadcasts during the keep-alive silence-fill window
+    /// after stdin reaches EOF. Default `.silence` preserves the historical
+    /// behavior (zero bytes, optionally TPDF-dithered). `.tone` substitutes a
+    /// continuous sine wave so listeners hear an audible "test tone" placeholder
+    /// instead of dead air.
+    var fillerMode: FillerMode = .silence
+    /// Frequency in Hz of the sine wave emitted when `fillerMode == .tone`.
+    /// Ignored otherwise. Default 1000 Hz is the broadcast convention for a
+    /// reference test tone.
+    var fillerToneHz: Double = 1000.0
+    /// Milliseconds of consecutive UDP/TCP input absence before the filler
+    /// kicks in. Lets brief network jitter pass through unaltered while still
+    /// covering longer gaps (Gqrx paused, station between feeds, etc.).
+    /// Default 500 ms matches the silence-dither threshold.
+    var fillerAfterMs: Int = 500
     var inputSource: PCMInputSource = .stdin
     var mountMP3: String = "/stream.mp3"
     var mountM4A: String = "/stream.m4a"
@@ -125,6 +140,24 @@ struct ServerConfig {
     var stdinChunkBytes: Int { stdinChunkFrames * bytesPerFrame }
 }
 
+// MARK: - Filler Mode
+
+/// Content the silence-fill loop emits when input has ended (stdin EOF + `--keep-alive`).
+enum FillerMode: String {
+    /// Continue the historical behavior: emit all-zero PCM (optionally TPDF-dithered).
+    case silence
+    /// Emit a continuous sine wave so listeners hear an audible placeholder.
+    case tone
+
+    init?(cliArgument: String) {
+        switch cliArgument.lowercased() {
+        case "silence":                    self = .silence
+        case "tone", "sine", "sine-tone":  self = .tone
+        default:                           return nil
+        }
+    }
+}
+
 // MARK: - Encoded Chunk
 
 /// A timestamped chunk of encoded audio bytes for one format.

Sources/LiveAudioServer/ConfigFile.swift

@@ -25,6 +25,7 @@ enum ConfigFileError: Error, CustomStringConvertible {
     case fileNotReadable(String)
     case decodeFailed(String, message: String)
     case invalidOutputs(String)
+    case invalidFillerMode(String)
 
     var description: String {
         switch self {
@@ -34,6 +35,8 @@ enum ConfigFileError: Error, CustomStringConvertible {
             return "Config file at \(p) is not valid JSON: \(m)"
         case .invalidOutputs(let token):
             return "Invalid value in config 'outputs': '\(token)'. Valid: mp3, aac, hls"
+        case .invalidFillerMode(let token):
+            return "Invalid value in config 'fillerMode': '\(token)'. Valid: silence, tone"
         }
     }
 }
@@ -64,6 +67,9 @@ struct ServerConfigFile: Codable {
     var reopenFIFO: Bool?
     var silenceDither: Bool?
     var silenceDitherMs: Int?
+    var fillerMode: String?       // "silence" | "tone"
+    var fillerToneHz: Double?
+    var fillerAfterMs: Int?
     var verbose: Bool?
     var bonjour: String?
     var bonjourInputs: Bool?
@@ -113,6 +119,14 @@ func applyConfigFile(_ file: ServerConfigFile, to config: inout ServerConfig) th
         // case via --silence-dither-ms.
         config.silenceDitherThresholdSamples = (v * config.sampleRate * config.channels) / 1000
     }
+    if let v = file.fillerMode {
+        guard let mode = FillerMode(cliArgument: v) else {
+            throw ConfigFileError.invalidFillerMode(v)
+        }
+        config.fillerMode = mode
+    }
+    if let v = file.fillerToneHz { config.fillerToneHz = v }
+    if let v = file.fillerAfterMs { config.fillerAfterMs = v }
     if let v = file.verbose       { config.verbose = v }
     if let v = file.bonjour       { config.bonjourName = v.isEmpty ? nil : v }
     if let v = file.bonjourInputs { config.bonjourAdvertiseInputs = v }

Sources/LiveAudioServer/LiveAudioServerApp.swift

@@ -128,6 +128,21 @@ func printUsage() {
                                 tools from treating the stream as dead.
       --silence-dither-ms <n>   Milliseconds of pure-silence before dither
                                 kicks in (default: 500).
+      --filler-mode <mode>      What to broadcast during the keep-alive
+                                silence-fill window after stdin EOF.
+                                "silence" (default) emits digital zero
+                                (optionally TPDF-dithered via
+                                --silence-dither). "tone" emits a continuous
+                                sine wave at --filler-tone-hz so listeners
+                                hear an audible "dead-air" placeholder.
+                                Only effective with --keep-alive.
+      --filler-tone-hz <hz>     Frequency of the sine-tone filler in Hz
+                                (default: 1000). Ignored unless
+                                --filler-mode tone.
+      --filler-after-ms <n>     Milliseconds of consecutive UDP/TCP input
+                                absence before the filler kicks in (default:
+                                500). Brief network jitter passes through
+                                unaltered; longer gaps switch to filler.
       -V, --verbose             Verbose logging
       -v, --version             Print version string and exit
       -h, --help                Show this help
@@ -391,6 +406,25 @@ func parseCLI(_ args: [String]) -> CLIParseResult {
                 return .error("Bad --silence-dither-ms (must be a non-negative integer)")
             }
             silenceDitherMs = ms
+        case "--filler-mode":
+            i += 1
+            guard i < args.count else { return .error("Missing --filler-mode value (silence|tone)") }
+            guard let mode = FillerMode(cliArgument: args[i]) else {
+                return .error("Bad --filler-mode '\(args[i])' (must be 'silence' or 'tone')")
+            }
+            config.fillerMode = mode
+        case "--filler-tone-hz":
+            i += 1
+            guard i < args.count, let v = Double(args[i]), v > 0, v < Double(config.sampleRate) / 2.0 else {
+                return .error("Bad --filler-tone-hz (must be > 0 and below the Nyquist frequency)")
+            }
+            config.fillerToneHz = v
+        case "--filler-after-ms":
+            i += 1
+            guard i < args.count, let v = Int(args[i]), v >= 0 else {
+                return .error("Bad --filler-after-ms (must be a non-negative integer)")
+            }
+            config.fillerAfterMs = v
         case "-V", "--verbose":
             config.verbose = true
         case "-v", "--version":