Add local spoken passages with Google retrieval and isolated runtime setup

crates/searchi-cli/src/main.rs

@@ -31,6 +31,9 @@ enum Command {
     },
     /// Set up the required Google model in the default local cache.
     Setup {
+        /// Set up the local speech recognition models instead of Google embeddings.
+        #[arg(long)]
+        speech: bool,
         /// Override the platform-local model cache directory.
         #[arg(long)]
         model_dir: Option<PathBuf>,
@@ -122,6 +125,12 @@ enum Command {
         /// Explicitly build only document vectors, without media.
         #[arg(long)]
         text_only: bool,
+        /// Optional bundled local speech worker; recognizes passages without replacing Google.
+        #[arg(long, requires = "speech_model_dir", conflicts_with = "text_only")]
+        speech_worker: Option<PathBuf>,
+        /// Directory containing verified base and Silero speech models.
+        #[arg(long, requires = "speech_worker", conflicts_with = "text_only")]
+        speech_model_dir: Option<PathBuf>,
     },
     /// Show the indexed file and content counts.
     Stats {
@@ -143,7 +152,17 @@ fn main() -> Result<()> {
             require_existing_index(&db)?;
             mcp::serve(db, model_dir, embedding_worker, usize::from(cpu_threads))?;
         }
-        Command::Setup { model_dir } => {
+        Command::Setup { model_dir, speech } => {
+            if speech {
+                let directory = model_dir
+                    .map(Ok)
+                    .unwrap_or_else(searchi_assets::speech::default_model_directory)?;
+                let progress = std::sync::atomic::AtomicU64::new(0);
+                eprintln!("Setting up local speech models; verified cached assets are reused (149 MB maximum download).");
+                searchi_assets::speech::provision(&directory, &progress)?;
+                println!("Speech models ready: {}", directory.display());
+                return Ok(());
+            }
             let directory = model_dir
                 .map(Ok)
                 .unwrap_or_else(searchi_assets::default_model_directory)?;
@@ -280,10 +299,21 @@ fn main() -> Result<()> {
             ffmpeg,
             ffprobe,
             text_only,
+            speech_worker,
+            speech_model_dir,
         } => {
             require_existing_index(&db)?;
             let mut service =
                 HybridSearch::open(&db, model_dir, embedding_worker, usize::from(cpu_threads))?;
+            if let Some((worker, directory)) = speech_worker.zip(speech_model_dir) {
+                let speech = searchi_speech::SpeechWorker::load(
+                    &worker,
+                    &directory.join("ggml-base.bin"),
+                    &directory.join("ggml-silero-v6.2.0.bin"),
+                    searchi_speech::Device::Auto,
+                )?;
+                service = service.with_speech_provider(Box::new(speech))?;
+            }
             let report = if text_only {
                 service.index_embeddings()
             } else {
@@ -353,6 +383,40 @@ fn require_existing_index(db: &std::path::Path) -> Result<()> {
 mod tests {
     use super::*;
 
+    #[test]
+    fn speech_indexing_requires_models_and_cannot_downgrade_to_text_only() {
+        let base = [
+            "searchi",
+            "embed",
+            "--db",
+            "index.db",
+            "--model-dir",
+            "google",
+            "--embedding-worker",
+            "google-worker",
+        ];
+        assert!(
+            Cli::try_parse_from(base.into_iter().chain(["--speech-worker", "speech-worker"]))
+                .is_err()
+        );
+        assert!(Cli::try_parse_from(base.into_iter().chain([
+            "--speech-worker",
+            "speech-worker",
+            "--speech-model-dir",
+            "speech",
+            "--text-only"
+        ]))
+        .is_err());
+        assert!(Cli::try_parse_from(base.into_iter().chain([
+            "--speech-worker",
+            "speech-worker",
+            "--speech-model-dir",
+            "speech"
+        ]))
+        .is_ok());
+        assert!(Cli::try_parse_from(["searchi", "setup", "--speech"]).is_ok());
+    }
+
     #[test]
     fn rejects_zero_or_excessive_result_limits() {
         for limit in ["0", "1001"] {