Skip to content

The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.

fix(speech): restore host path context in transcription errors

Steven Enamakel0583cad171b6
diff --git a/crates/openhuman-core/src/inference/host_runtime/service/speech.rs b/crates/openhuman-core/src/inference/host_runtime/service/speech.rs
index f4fa7d1a88..0199718a97 100644
--- a/crates/openhuman-core/src/inference/host_runtime/service/speech.rs
+++ b/crates/openhuman-core/src/inference/host_runtime/service/speech.rs
@@ -51,9 +51,9 @@ pub async fn transcribe_with_prompt(
let provider_name = effective_stt_provider(config);
debug!("{LOG_PREFIX} configured STT dispatch provider={provider_name} path={audio_path}");
let provider = selected_stt_provider(config, &provider_name);
- let text =
- tinyinference_voice::transcription::transcribe_file(Path::new(audio_path), &provider)
- .await?;
+ let text = tinyinference_voice::transcription::transcribe_file(Path::new(audio_path), &provider)
+ .await
+ .map_err(|error| restore_host_file_error(error, audio_path))?;
debug!(
"{LOG_PREFIX} configured STT complete (provider={} elapsed_ms={})",
provider_name,
@@ -66,6 +66,16 @@ pub async fn transcribe_with_prompt(
})
}
+fn restore_host_file_error(error: String, audio_path: &str) -> String {
+ if let Some(reason) = error.strip_prefix("[voice-stt] failed to read audio file: ") {
+ return format!("failed to read audio file {audio_path}: {reason}");
+ }
+ if error == "[voice-stt] audio file is empty" {
+ return format!("audio file {audio_path} is empty");
+ }
+ error
+}
+
/// Transcribe audio bytes with the host-selected STT provider.
pub async fn transcribe_bytes(
service: &LocalAiService,
diff --git a/crates/openhuman-core/src/inference/host_runtime/service/speech_tests.rs b/crates/openhuman-core/src/inference/host_runtime/service/speech_tests.rs
index e34d6240ab..2c257f8915 100644
--- a/crates/openhuman-core/src/inference/host_runtime/service/speech_tests.rs
+++ b/crates/openhuman-core/src/inference/host_runtime/service/speech_tests.rs
@@ -5,6 +5,26 @@ use async_trait::async_trait;
use base64::{engine::general_purpose::STANDARD as BASE64, Engine};
use std::sync::{Arc, Mutex};
+#[test]
+fn file_transcription_errors_keep_the_host_path_context() {
+ let missing = restore_host_file_error(
+ "[voice-stt] failed to read audio file: No such file or directory (os error 2)".into(),
+ "/tmp/missing.wav",
+ );
+ assert_eq!(
+ missing,
+ "failed to read audio file /tmp/missing.wav: No such file or directory (os error 2)"
+ );
+
+ let empty = restore_host_file_error("[voice-stt] audio file is empty".into(), "/tmp/empty.wav");
+ assert_eq!(empty, "audio file /tmp/empty.wav is empty");
+
+ assert_eq!(
+ restore_host_file_error("provider refused audio".into(), "/tmp/audio.wav"),
+ "provider refused audio"
+ );
+}
+
#[derive(Debug, PartialEq, Eq)]
struct CapturedAudio {
bytes: Vec<u8>,