Added TTS and ASR module

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "dialogflow"
-version = "1.16.1"
+version = "1.17.0"
 edition = "2021"
 homepage = "https://dialogflowchatbot.github.io/"
 authors = ["dialogflowchatbot <[email protected]>"]
@@ -18,10 +18,12 @@ anyhow = "1.0"
 axum = {version = "0.7", features = ["query", "tokio", "macros"]}
 bigdecimal = "0.4"
 # candle-core = { git = "https://github.com/huggingface/candle.git", version = "0.5.1" }
-candle = { version = "0.6", package = "candle-core", default-features = false }
-candle-nn = "0.6"
+# candle = { version = "0.6", package = "candle-core", default-features = false }
+candle = { git = "https://github.com/huggingface/candle.git", package = "candle-core", default-features = false }
+candle-nn = { git = "https://github.com/huggingface/candle.git" }
 # candle-onnx = "0.6"
-candle-transformers = { version = "0.6" }
+candle-transformers = { git = "https://github.com/huggingface/candle.git" }
+# candle-transformers = { version = "0.6" }
 # candle-transformers = { version = "0.6", features = ["flash-attn"] }
 # crossbeam-channel = "0.5"
 frand = "0.10"

src/ai/asr.rs

@@ -0,0 +1,9 @@
+use serde::{Deserialize, Serialize};
+
+use super::huggingface::{load_bert_model_files, HuggingFaceModel, HuggingFaceModelInfo};
+
+#[derive(Clone, Deserialize, Serialize)]
+#[serde(tag = "id", content = "model")]
+pub(crate) enum AsrProvider {
+    HuggingFace(HuggingFaceModel),
+}

src/ai/audio.rs

@@ -0,0 +1,31 @@
+use candle::Tensor;
+
+use crate::result::Result;
+
+pub fn normalize_loudness(
+    wav: &Tensor,
+    sample_rate: u32,
+    loudness_compressor: bool,
+) -> Result<Tensor> {
+    let energy = wav.sqr()?.mean_all()?.sqrt()?.to_vec0::<f32>()?;
+    if energy < 2e-3 {
+        return Ok(wav.clone());
+    }
+    let wav_array = wav.to_vec1::<f32>()?;
+    let mut meter = super::bs1770::ChannelLoudnessMeter::new(sample_rate);
+    meter.push(wav_array.into_iter());
+    let power = meter.as_100ms_windows();
+    let loudness = match super::bs1770::gated_mean(power) {
+        None => return Ok(wav.clone()),
+        Some(gp) => gp.loudness_lkfs() as f64,
+    };
+    let delta_loudness = -14. - loudness;
+    let gain = 10f64.powf(delta_loudness / 20.);
+    let wav = (wav * gain)?;
+    if loudness_compressor {
+        let r = wav.tanh()?;
+        Ok(r)
+    } else {
+        Ok(wav)
+    }
+}