langchain_rust/tools/text2speech/openai/
client.rs

1use std::{error::Error, sync::Arc};
2
3use async_openai::types::audio::CreateSpeechRequestArgs;
4use async_openai::Client;
5pub use async_openai::{
6    config::{Config, OpenAIConfig},
7    types::audio::{SpeechModel, SpeechResponseFormat, Voice},
8};
9use async_trait::async_trait;
10use serde_json::Value;
11
12use crate::tools::{SpeechStorage, Tool};
13
14#[derive(Clone)]
15pub struct Text2SpeechOpenAI<C: Config> {
16    config: C,
17    model: SpeechModel,
18    voice: Voice,
19    storage: Option<Arc<dyn SpeechStorage>>,
20    response_format: SpeechResponseFormat,
21    path: String,
22}
23
24impl<C: Config> Text2SpeechOpenAI<C> {
25    pub fn new(config: C) -> Self {
26        Self {
27            config,
28            model: SpeechModel::Tts1,
29            voice: Voice::Alloy,
30            storage: None,
31            response_format: SpeechResponseFormat::Mp3,
32            path: "./data/audio.mp3".to_string(),
33        }
34    }
35
36    pub fn with_model(mut self, model: SpeechModel) -> Self {
37        self.model = model;
38        self
39    }
40
41    pub fn with_voice(mut self, voice: Voice) -> Self {
42        self.voice = voice;
43        self
44    }
45
46    pub fn with_storage<SS: SpeechStorage + 'static>(mut self, storage: SS) -> Self {
47        self.storage = Some(Arc::new(storage));
48        self
49    }
50
51    pub fn with_response_format(mut self, response_format: SpeechResponseFormat) -> Self {
52        self.response_format = response_format;
53        self
54    }
55
56    pub fn with_path<S: Into<String>>(mut self, path: S) -> Self {
57        self.path = path.into();
58        self
59    }
60
61    pub fn with_config(mut self, config: C) -> Self {
62        self.config = config;
63        self
64    }
65}
66
67impl Default for Text2SpeechOpenAI<OpenAIConfig> {
68    fn default() -> Self {
69        Self::new(OpenAIConfig::default())
70    }
71}
72
73#[async_trait]
74impl<C: Config + Clone + Send + Sync> Tool for Text2SpeechOpenAI<C> {
75    fn name(&self) -> String {
76        "Text2SpeechOpenAI".to_string()
77    }
78
79    fn description(&self) -> String {
80        r#"A wrapper around OpenAI Text2Speech. "
81        "Useful for when you need to convert text to speech. "
82        "It supports multiple languages, including English, German, Polish, "
83        "Spanish, Italian, French, Portuguese""#
84            .to_string()
85    }
86
87    async fn run(&self, input: Value) -> Result<String, Box<dyn Error>> {
88        let input = input.as_str().ok_or("Invalid input")?;
89        let client = Client::with_config(self.config.clone());
90        let response_format: SpeechResponseFormat = self.response_format;
91
92        let request = CreateSpeechRequestArgs::default()
93            .input(input)
94            .voice(self.voice.clone())
95            .response_format(response_format)
96            .model(self.model.clone())
97            .build()?;
98
99        let response = client.audio().speech().create(request).await?;
100
101        if self.storage.is_some() {
102            let storage = self.storage.as_ref().unwrap(); //safe to unwrap
103            let data = response.bytes;
104            return storage.save(&self.path, &data).await;
105        } else {
106            response.save(&self.path).await?;
107        }
108
109        Ok(self.path.clone())
110    }
111}
112
113#[cfg(test)]
114mod tests {
115    use crate::tools::{Text2SpeechOpenAI, Tool};
116
117    #[tokio::test]
118    #[ignore]
119    async fn openai_speech2text_tool() {
120        let openai = Text2SpeechOpenAI::default();
121        let s = openai.call("Hola como estas").await.unwrap();
122        println!("{}", s);
123    }
124
125    #[tokio::test]
126    async fn speech_uses_configured_client_and_writes_file() {
127        use super::*;
128        use serde_json::json;
129
130        let mut server = mockito::Server::new_async().await;
131        let mock = server
132            .mock("POST", "/audio/speech")
133            .match_header("authorization", "Bearer test-key")
134            .match_body(mockito::Matcher::PartialJson(json!({
135                "input": "hola",
136                "model": "tts-1-hd",
137                "voice": "nova",
138                "response_format": "wav"
139            })))
140            .with_body([1u8, 2, 3])
141            .create_async()
142            .await;
143
144        let nanos = std::time::SystemTime::now()
145            .duration_since(std::time::UNIX_EPOCH)
146            .unwrap()
147            .as_nanos();
148        let dir = std::env::temp_dir().join(format!("lcr-tts-{}-{nanos}", std::process::id()));
149        let path = dir.join("out.wav");
150        let tool = Text2SpeechOpenAI::new(
151            OpenAIConfig::new()
152                .with_api_base(server.url())
153                .with_api_key("test-key"),
154        )
155        .with_model(SpeechModel::Tts1Hd)
156        .with_voice(Voice::Nova)
157        .with_response_format(SpeechResponseFormat::Wav)
158        .with_path(path.to_string_lossy());
159
160        let out = tool.run(json!("hola")).await.unwrap();
161        mock.assert_async().await;
162        assert_eq!(out, path.to_string_lossy());
163        assert_eq!(std::fs::read(&path).unwrap(), vec![1, 2, 3]);
164        let _ = std::fs::remove_dir_all(dir);
165    }
166}
langchain_rust/tools/text2speech/openai/client.rs

langchain_rust/tools/text2speech/openai/
client.rs
