feat(plugins): add Parakeet TDT speech-to-text plugin

Add a new native plugin for fast English speech recognition using NVIDIA's
Parakeet TDT (Token-and-Duration Transducer) 0.6B model via sherpa-onnx.

Parakeet TDT is approximately 10x faster than Whisper on consumer hardware
with competitive accuracy (#1 on HuggingFace ASR leaderboard).

Plugin implementation:
- Offline transducer recognizer (encoder/decoder/joiner) via sherpa-onnx C API
- Silero VAD v6 for streaming speech segmentation
- Recognizer caching keyed on (model_dir, num_threads, execution_provider)
- Configurable VAD threshold, silence duration, and max segment length
- 16kHz mono f32 audio input, transcription output

Justfile additions:
- build-plugin-native-parakeet: build the plugin
- download-parakeet-models: download INT8 quantized model (~660MB)
- setup-parakeet: full setup (sherpa-onnx + models + VAD)
- Added parakeet to copy-plugins-native loop

Includes sample oneshot pipeline (parakeet-stt.yml) and plugin.yml manifest.

Signed-off-by: StreamKit Devin <devin@streamkit.dev>
Co-Authored-By: Claudio Costa <cstcld91@gmail.com>

Author: streamkit-devin

justfile

@@ -723,6 +723,43 @@ upload-sensevoice-plugin: build-plugin-native-sensevoice
     @curl -X POST -F "plugin=@{{plugins_target_dir}}/release/libsensevoice.so" \
         http://127.0.0.1:4545/api/v1/plugins
 
+# Build native Parakeet TDT STT plugin
+[working-directory: 'plugins/native/parakeet']
+build-plugin-native-parakeet:
+    @echo "Building native Parakeet TDT STT plugin..."
+    @CARGO_TARGET_DIR={{plugins_target_dir}} cargo build --release
+
+# Upload Parakeet plugin to running server
+[working-directory: 'plugins/native/parakeet']
+upload-parakeet-plugin: build-plugin-native-parakeet
+    @echo "Uploading Parakeet plugin to server..."
+    @curl -X POST -F "plugin=@{{plugins_target_dir}}/release/libparakeet.so" \
+        http://127.0.0.1:4545/api/v1/plugins
+
+# Download Parakeet TDT models
+download-parakeet-models:
+    @echo "Downloading Parakeet TDT models..."
+    @mkdir -p models
+    @if [ -f models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2 ]; then \
+        echo "✓ Parakeet TDT archive already exists"; \
+    else \
+        echo "Downloading sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2 (~660MB)..." && \
+        curl -L -o models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2 \
+            https://huggingface.co/csukuangfj/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/resolve/main/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2 && \
+        echo "✓ Parakeet TDT archive downloaded"; \
+    fi
+    @if [ -d models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8 ]; then \
+        echo "✓ Parakeet TDT models already extracted"; \
+    else \
+        echo "Extracting models..." && \
+        cd models && tar xf sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2 && \
+        echo "✓ Parakeet TDT models ready at models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8 (English)"; \
+    fi
+
+# Setup Parakeet (install dependencies + download models)
+setup-parakeet: install-sherpa-onnx download-parakeet-models download-silero-vad
+    @echo "✓ Parakeet TDT STT setup complete!"
+
 # Download pre-converted NLLB models from Hugging Face
 download-nllb-models:
     @echo "Downloading pre-converted NLLB-200 models from Hugging Face..."
@@ -1042,7 +1079,7 @@ copy-plugins-native:
 
     # Official native plugins (shared target dir).
     # For most plugins the lib stem matches the plugin id.
-    for name in whisper kokoro piper matcha vad sensevoice nllb helsinki supertonic slint; do
+    for name in whisper kokoro piper matcha vad sensevoice nllb helsinki supertonic slint parakeet; do
         copy_plugin "$name" "$name" "$PLUGINS_TARGET"
     done