Add: Azure speech cognitive service - get transcription report (v3.1)

examples/azure/Cargo.toml

@@ -14,5 +14,5 @@ name = "azure-text-to-speech"
 path = "src/text-to-speech.rs"
 
 [[bin]]
-name = "azure-batch-create-transcription"
-path = "src/batch-create-transcription.rs"
+name = "azure-batch-transcription"
+path = "src/batch-transcription.rs"

examples/azure/src/batch-create-transcription.rs → examples/azure/src/batch-transcription.rs

@@ -15,13 +15,22 @@ async fn main() -> Result<(), Box<dyn Error>> {
         .await?;
 
     std::thread::sleep(Duration::from_secs(5));
+
+    // Check transcription job status.
     let trans = trans.status().await?;
     if let Some(Status::Succeeded) = trans.status {
+        // Get transcription result files.
         let results = trans.files().await?;
-        let files = results.values;
+        let files = results.values.clone();
 
         if files.len() > 0 {
-            println!("{:#?}", files.get(0).unwrap().file().await?);
+            // Get transcription report.
+            let report = results.report().await?;
+            println!("{:#?}", report);
+
+            // Get transcription result file via individual API endpoint.
+            let file = files.get(0).unwrap().file().await?;
+            println!("{:#?}", file);
         }
     }
 

rust-ai/src/azure/apis/speech.rs

@@ -43,10 +43,10 @@ use crate::azure::{
         common::{MicrosoftOutputFormat, ResponseExpectation, ResponseType},
         speech::{
             entity::EntityReference,
-            file::{File, PaginatedFiles},
+            file::{File, FileKind, PaginatedFiles},
             filter::FilterOperator,
             health::ServiceHealth,
-            transcription::{Status, Transcription},
+            transcription::{Status, Transcription, TranscriptionReport},
             ErrorResponse,
         },
         tts::Voice,
@@ -414,6 +414,46 @@ impl PaginatedFiles {
             Ok(None)
         }
     }
+
+    pub async fn report(self) -> Result<Option<TranscriptionReport>, Box<dyn std::error::Error>> {
+        if self.values.len() == 0 {
+            return Ok(None);
+        }
+
+        let mut report = self
+            .values
+            .iter()
+            .filter(|file| file.kind == FileKind::TranscriptionReport);
+
+        if let Some(report) = report.next().clone() {
+            let text = request_get_endpoint(
+                &SpeechServiceEndpoint::None,
+                None,
+                Some(report.links.content_url.clone()),
+            )
+            .await?;
+
+            return match serde_json::from_str::<TranscriptionReport>(&text) {
+                Ok(report) => Ok(Some(report)),
+                Err(e) => {
+                    warn!(target: "azure", "Unable to parse transcription result file: `{:#?}`", e);
+                    match serde_json::from_str::<ErrorResponse>(&text) {
+                        Ok(error) => {
+                            println!("{:#?}", error);
+                            error!(target: "azure", "Error from Azure: `{:?}`", e);
+                            Err(Box::new(e))
+                        }
+                        Err(e) => {
+                            error!(target: "azure", "Unable to parse error response: `{:?}`", e);
+                            Err(Box::new(e))
+                        }
+                    }
+                }
+            };
+        } else {
+            Ok(None)
+        }
+    }
 }
 
 impl File {

rust-ai/src/azure/endpoint.rs

@@ -22,6 +22,7 @@ pub enum SpeechServiceEndpoint {
     Get_Transcription_v3_1,
     Get_Transcription_Files_v3_1,
     Get_Transcription_File_v3_1,
+    None,
 }
 
 impl SpeechServiceEndpoint {
@@ -56,20 +57,24 @@ impl SpeechServiceEndpoint {
                 "https://{}.api.cognitive.microsoft.com/speechtotext/v3.1/transcriptions/",
                 region
             ),
-            
+
             Self::Get_Transcription_Files_v3_1 => format!(
                 "https://{}.api.cognitive.microsoft.com/speechtotext/v3.1/transcriptions/",
                 region
             ),
-            
+
             Self::Get_Transcription_File_v3_1 => format!(
                 "https://{}.api.cognitive.microsoft.com/speechtotext/v3.1/transcriptions/",
                 region
             ),
+            Self::None => String::new(),
         }
     }
 }
 
+/// If you would like to use a plain request URI, pass 
+/// [`SpeechServiceEndpoint::None`] as the endpoint, and then provide the 
+/// complete URI through `url_suffix` parameter.
 pub async fn request_get_endpoint(
     endpoint: &SpeechServiceEndpoint,
     params: Option<HashMap<String, String>>,
@@ -78,7 +83,10 @@ pub async fn request_get_endpoint(
     let config = Config::load().unwrap();
     let region = config.azure.speech.region;
 
-    let mut url = endpoint.build(&region);
+    let mut url = match endpoint {
+        SpeechServiceEndpoint::None => String::new(),
+        _ => endpoint.build(&region),
+    };
 
     if let Some(url_suffix) = url_suffix {
         url.push_str(&url_suffix);

rust-ai/src/azure/types/speech/file.rs

@@ -74,7 +74,7 @@ impl File {
 }
 
 /// Type of data.
-#[derive(Clone, Debug, Deserialize, Serialize)]
+#[derive(Clone, Debug, Deserialize, Serialize, PartialEq, Eq)]
 pub enum FileKind {
     /// Type of data is dataset report.
     DatasetReport,

rust-ai/src/azure/types/speech/transcription.rs

@@ -355,3 +355,20 @@ impl Default for TranscriptionProperties {
         }
     }
 }
+
+#[derive(Clone, Debug, Deserialize, Serialize)]
+pub struct TranscriptionReport {
+    #[serde(rename = "successfulTranscriptionsCount")]
+    pub successful_transcriptions_count: usize,
+
+    #[serde(rename = "failedTranscriptionsCount")]
+
+    pub failed_transcriptions_count: usize,
+    pub details: Vec<TranscriptionReportDetail>,
+}
+
+#[derive(Clone, Debug, Deserialize, Serialize)]
+pub struct TranscriptionReportDetail {
+    pub source: String,
+    pub status: String,
+}