WIP

livepeer · Jul 25, 2024 · 3072fcb · 3072fcb
1 parent aeb257f
commit 3072fcb
Show file tree

Hide file tree

Showing 4 changed files with 51 additions and 73 deletions.
diff --git a/runner/app/routes/text_to_audio.py b/runner/app/routes/text_to_audio.py
@@ -26,24 +26,33 @@ class HTTPError(BaseModel):
     }
 }
 
-@router.post("/text-to-audio", responses=responses)
+@router.post("/text-to-audio",
+    summary="Text to Audio",
+    operation_id="text_to_audio",
+    responses=responses)
 async def TextToAudio(
-    text_input: Optional[str] = Form(None),
+    text_input: str = Form(...),
     pipeline: Pipeline = Depends(get_pipeline),
 ):
 
     try:
-        result = pipeline(
-            text_input
+        if not text_input:
+            raise ValueError("text_input is required and cannot be empty.")
+
+        result = pipeline(text_input)
+
+    except ValueError as ve:
+        logger.error(f"Validation error: {ve}")
+        return JSONResponse(
+            status_code=400,
+            content={"detail": str(ve)},
         )
 
     except Exception as e:
         logger.error(f"TextToAudioPipeline error: {e}")
         return JSONResponse(
             status_code=500,
-            content={
-                "detail": f"Internal Server Error: {str(e)}"
-            },
+            content={"detail": f"Internal Server Error: {str(e)}"},
         )
 
     if os.path.exists(result):

diff --git a/runner/gen_openapi.py b/runner/gen_openapi.py
@@ -13,6 +13,7 @@
     image_to_video,
     text_to_image,
     upscale,
+    text_to_audio,
 )
 from fastapi.openapi.utils import get_openapi
 

diff --git a/runner/openapi.json b/runner/openapi.json
@@ -407,21 +407,17 @@
         },
         "/text-to-audio": {
             "post": {
-                "summary": "Texttoaudio",
+                "summary": "Text to Audio",
                 "operationId": "TextToAudio",
                 "requestBody": {
                     "content": {
                         "application/x-www-form-urlencoded": {
                             "schema": {
-                                "allOf": [
-                                    {
-                                        "$ref": "#/components/schemas/Body_TextToAudio_text_to_audio_post"
-                                    }
-                                ],
-                                "title": "Body"
+                                "$ref": "#/components/schemas/Body_text_to_audio"
                             }
                         }
-                    }
+                    },
+                    "required": true
                 },
                 "responses": {
                     "200": {
@@ -502,22 +498,24 @@
                 ],
                 "title": "APIError"
             },
-            "Body_TextToAudio_text_to_audio_post": {
+            "Body_audio_to_text_audio_to_text_post": {
                 "properties": {
-                    "text_input": {
-                        "anyOf": [
-                            {
-                                "type": "string"
-                            },
-                            {
-                                "type": "null"
-                            }
-                        ],
-                        "title": "Text Input"
+                    "audio": {
+                        "type": "string",
+                        "format": "binary",
+                        "title": "Audio"
+                    },
+                    "model_id": {
+                        "type": "string",
+                        "title": "Model Id",
+                        "default": ""
                     }
                 },
                 "type": "object",
-                "title": "Body_TextToAudio_text_to_audio_post"
+                "required": [
+                    "audio"
+                ],
+                "title": "Body_audio_to_text_audio_to_text_post"
             },
             "Body_image_to_image_image_to_image_post": {
                 "properties": {
@@ -640,6 +638,19 @@
                 ],
                 "title": "Body_image_to_video_image_to_video_post"
             },
+            "Body_text_to_audio": {
+                "properties": {
+                    "text_input": {
+                        "type": "string",
+                        "title": "Text Input"
+                    }
+                },
+                "type": "object",
+                "required": [
+                    "text_input"
+                ],
+                "title": "Body_text_to_audio"
+            },
             "Body_upscale_upscale_post": {
                 "properties": {
                     "prompt": {

diff --git a/worker/worker.go b/worker/worker.go
@@ -251,57 +251,14 @@ func (w *Worker) Upscale(ctx context.Context, req UpscaleMultipartRequestBody) (
 }
 
 
-func (w *Worker) AudioToText(ctx context.Context, req TextToAudioJSONRequestBody) (*AudioResponse, error) {
-	c, err := w.borrowContainer(ctx, "audio-to-text", *req.ModelId)
-	if err != nil {
-		return nil, err
-	}
-	defer w.returnContainer(c)
-
-	resp, err := c.Client.AudioToTextWithResponse(ctx, req)
-	if err != nil {
-		return nil, err
-	}
-
-	if resp.JSON422 != nil {
-		val, err := json.Marshal(resp.JSON422)
-		if err != nil {
-			return nil, err
-		}
-		slog.Error("audio-to-text container returned 422", slog.String("err", string(val)))
-		return nil, errors.New("audio-to-text container returned 422")
-	}
-
-	if resp.JSON400 != nil {
-		val, err := json.Marshal(resp.JSON400)
-		if err != nil {
-			return nil, err
-		}
-		slog.Error("audio-to-text container returned 400", slog.String("err", string(val)))
-		return nil, errors.New("audio-to-text container returned 400")
-	}
-
-	if resp.JSON500 != nil {
-		val, err := json.Marshal(resp.JSON500)
-		if err != nil {
-			return nil, err
-		}
-		slog.Error("audio-to-text container returned 500", slog.String("err", string(val)))
-		return nil, errors.New("audio-to-text container returned 500")
-	}
-
-	return resp.JSON200, nil
-}
-
-
-func (w *Worker) AudioToText(ctx context.Context, req TextToAudioJSONRequestBody) (*AudioResponse, error) {
+func (w *Worker) TextToAudio(ctx context.Context, req TextToAudioJSONRequestBody) (*AudioResponse, error) {
 	c, err := w.borrowContainer(ctx, "text-to-audio", *req.ModelId)
 	if err != nil {
 		return nil, err
 	}
 	defer w.returnContainer(c)
 
-	resp, err := c.Client.AudioToTextWithResponse(ctx, req)
+	resp, err := c.Client.TextToAudioWithResponse(ctx, req)
 	if err != nil {
 		return nil, err
 	}
@@ -330,7 +287,7 @@ func (w *Worker) AudioToText(ctx context.Context, req TextToAudioJSONRequestBody
 			return nil, err
 		}
 		slog.Error("text-to-audio container returned 500", slog.String("err", string(val)))
-		return nil, errors.New("text-to-image container returned 500")
+		return nil, errors.New("text-to-audio container returned 500")
 	}
 
 	return resp.JSON200, nil