tckm.yan/aha
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(asr): add text response format support and improve code formatting

Browse Source
This commit is contained in:
XiaoYang
2026-03-06 12:00:39 +08:00
parent 63a89b0c96
commit b559376438
1 changed files with 32 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/api/asr.rs
+30 -12
View File
@@ -4,15 +4,15 @@
use aha::models::GenerateModel; use aha::models::GenerateModel;
use aha::utils::{clean_asr_response, map_language_code}; use aha::utils::{clean_asr_response, map_language_code};
use aha_openai_dive::v1::resources::chat::{ use aha_openai_dive::v1::resources::chat::{
ChatCompletionParameters, ChatMessage, ChatMessageAudioContentPart, ChatMessageContent, AudioUrlType, ChatCompletionParameters, ChatMessage, ChatMessageAudioContentPart,
ChatMessageContentPart, AudioUrlType, ChatMessageContent, ChatMessageContentPart,
}; };
use rocket::http::Status; use rocket::http::Status;
use rocket::serde::json::Json; use rocket::serde::json::Json;
use rocket::{form::Form, post}; use rocket::{form::Form, post};
use super::asr_types::{ErrorResponse, ErrorDetail, TranscriptionRequest, TranscriptionResponse};
use super::MODEL; use super::MODEL;
use super::asr_types::{ErrorDetail, ErrorResponse, TranscriptionRequest, TranscriptionResponse};
/// Handle audio transcription requests /// Handle audio transcription requests
/// ///
@@ -30,14 +30,17 @@ use super::MODEL;
/// # Returns /// # Returns
/// JSON response with format: `{"text": "transcribed text"}` /// JSON response with format: `{"text": "transcribed text"}`
#[post("/transcriptions", data = "<req>")] #[post("/transcriptions", data = "<req>")]
pub(crate) async fn transcriptions(req: Form<TranscriptionRequest<'_>>) -> (Status, Json<serde_json::Value>) { pub(crate) async fn transcriptions(
// Validate response_format (only JSON supported) req: Form<TranscriptionRequest<'_>>,
) -> (Status, Json<serde_json::Value>) {
// Validate response_format (only JSON and text supported)
#[allow(clippy::collapsible_if)]
if let Some(ref format) = req.response_format { if let Some(ref format) = req.response_format {
if format != "json" && format != "text" { if format != "json" && format != "text" {
return error_response( return error_response(
Status::BadRequest, Status::BadRequest,
"invalid_request_error", "invalid_request_error",
"Only 'json' response format is supported", "Only 'json' and 'text' response formats are supported",
Some("unsupported_format".to_string()), Some("unsupported_format".to_string()),
); );
} }
@@ -60,7 +63,10 @@ pub(crate) async fn transcriptions(req: Form<TranscriptionRequest<'_>>) -> (Stat
let file_url = format!("file://{}", file_path.display()); let file_url = format!("file://{}", file_path.display());
// Map language code to full language name // Map language code to full language name
let language_name = req.language.as_ref().and_then(|code| map_language_code(code)); let language_name = req
.language
.as_ref()
.and_then(|code| map_language_code(code));
// Build ChatCompletionParameters for the ASR model // Build ChatCompletionParameters for the ASR model
let audio_part = ChatMessageContentPart::Audio(ChatMessageAudioContentPart { let audio_part = ChatMessageContentPart::Audio(ChatMessageAudioContentPart {
@@ -130,22 +136,34 @@ pub(crate) async fn transcriptions(req: Form<TranscriptionRequest<'_>>) -> (Stat
None None
} }
}) })
.unwrap_or_else(|| String::new()); .unwrap_or_else(String::new);
// Clean the response (remove "language English<asr_text>" prefix) // Clean the response (remove "language English<asr_text>" prefix)
let cleaned_text = clean_asr_response(&raw_text); let cleaned_text = clean_asr_response(&raw_text);
// Return OpenAI-compatible transcription response // Return OpenAI-compatible transcription response
let transcription = TranscriptionResponse { text: cleaned_text }; let transcription = TranscriptionResponse { text: cleaned_text };
(Status::Ok, Json(serde_json::to_value(transcription).unwrap())) (
Status::Ok,
Json(serde_json::to_value(transcription).unwrap()),
)
} }
Err(e) => { Err(e) => {
// Determine appropriate error status based on error message // Determine appropriate error status based on error message
let error_msg = e.to_string(); let error_msg = e.to_string();
let (status, error_type, code) = if error_msg.contains("audio") || error_msg.contains("decode") { let (status, error_type, code) =
(Status::BadRequest, "invalid_request_error", Some("audio_decode_error".to_string())) if error_msg.contains("audio") || error_msg.contains("decode") {
(
Status::BadRequest,
"invalid_request_error",
Some("audio_decode_error".to_string()),
)
} else { } else {
(Status::InternalServerError, "server_error", Some("inference_error".to_string())) (
Status::InternalServerError,
"server_error",
Some("inference_error".to_string()),
)
}; };
error_response(status, error_type, &error_msg, code) error_response(status, error_type, &error_msg, code)