Commit @b29e0ba26b152110d52626a92e7f61fc51498ad8 - yjyoon/whisper_server

Fedir Zadniprovskyi 2024-06-14

chore: improve api docs

@b29e0ba26b152110d52626a92e7f61fc51498ad8

1f760da

b29e0ba

faster_whisper_server/main.py

--- faster_whisper_server/main.py

+++ faster_whisper_server/main.py


 
 
 @app.get("/v1/models/{model_name:path}")
-def get_model(model_name: Annotated[str, Path()]) -> ModelObject:
+# NOTE: `examples` doesn't work https://github.com/tiangolo/fastapi/discussions/10537
+def get_model(
+    model_name: Annotated[str, Path(example="Systran/faster-distil-whisper-large-v3")],
+) -> ModelObject:
     models = list(
         huggingface_hub.list_models(model_name=model_name, library="ctranslate2")
     )

 ModelName = Annotated[str, AfterValidator(handle_default_openai_model)]
 
 
-@app.post("/v1/audio/translations")
+@app.post(
+    "/v1/audio/translations",
+    response_model=str | TranscriptionJsonResponse | TranscriptionVerboseJsonResponse,
+)
 def translate_file(
     file: Annotated[UploadFile, Form()],
     model: Annotated[ModelName, Form()] = config.whisper.model,

     response_format: Annotated[ResponseFormat, Form()] = config.default_response_format,
     temperature: Annotated[float, Form()] = 0.0,
     stream: Annotated[bool, Form()] = False,
+) -> (
+    str
+    | TranscriptionJsonResponse
+    | TranscriptionVerboseJsonResponse
+    | StreamingResponse
 ):
     start = time.perf_counter()
     whisper = load_model(model)

 
 # https://platform.openai.com/docs/api-reference/audio/createTranscription
 # https://github.com/openai/openai-openapi/blob/master/openapi.yaml#L8915
-@app.post("/v1/audio/transcriptions")
+@app.post(
+    "/v1/audio/transcriptions",
+    response_model=str | TranscriptionJsonResponse | TranscriptionVerboseJsonResponse,
+)
 def transcribe_file(
     file: Annotated[UploadFile, Form()],
     model: Annotated[ModelName, Form()] = config.whisper.model,

         Form(alias="timestamp_granularities[]"),
     ] = ["segment"],
     stream: Annotated[bool, Form()] = False,
+) -> (
+    str
+    | TranscriptionJsonResponse
+    | TranscriptionVerboseJsonResponse
+    | StreamingResponse
 ):
     start = time.perf_counter()
     whisper = load_model(model)

1f760da

b29e0ba

faster_whisper_server/server_models.py

--- faster_whisper_server/server_models.py

+++ faster_whisper_server/server_models.py


 
 
 class ModelObject(BaseModel):
-    model_config = ConfigDict(populate_by_name=True)
-
     id: str
     """The model identifier, which can be referenced in the API endpoints."""
     created: int

     """The object type, which is always "model"."""
     owned_by: str
     """The organization that owns the model."""
+
+    model_config = ConfigDict(
+        populate_by_name=True,
+        json_schema_extra={
+            "examples": [
+                {
+                    "id": "Systran/faster-whisper-large-v3",
+                    "created": 1700732060,
+                    "object": "model",
+                    "owned_by": "Systran",
+                },
+                {
+                    "id": "Systran/faster-distil-whisper-large-v3",
+                    "created": 1711378296,
+                    "object": "model",
+                    "owned_by": "Systran",
+                },
+                {
+                    "id": "bofenghuang/whisper-large-v2-cv11-french-ct2",
+                    "created": 1687968011,
+                    "object": "model",
+                    "owned_by": "bofenghuang",
+                },
+            ]
+        },
+    )

Add a comment

Open 0
Closed 0

List

...	...	@@ -104,7 +104,10 @@
104	104
105	105
106	106	@app.get("/v1/models/{model_name:path}")
107		-def get_model(model_name: Annotated[str, Path()]) -> ModelObject:
	107	+# NOTE: `examples` doesn't work https://github.com/tiangolo/fastapi/discussions/10537
	108	+def get_model(
	109	+ model_name: Annotated[str, Path(example="Systran/faster-distil-whisper-large-v3")],
	110	+) -> ModelObject:
108	111	models = list(
109	112	huggingface_hub.list_models(model_name=model_name, library="ctranslate2")
110	113	)
...	...	@@ -148,7 +151,10 @@
148	151	ModelName = Annotated[str, AfterValidator(handle_default_openai_model)]
149	152
150	153
151		-@app.post("/v1/audio/translations")
	154	+@app.post(
	155	+ "/v1/audio/translations",
	156	+ response_model=str \| TranscriptionJsonResponse \| TranscriptionVerboseJsonResponse,
	157	+)
152	158	def translate_file(
153	159	file: Annotated[UploadFile, Form()],
154	160	model: Annotated[ModelName, Form()] = config.whisper.model,
...	...	@@ -156,6 +162,11 @@
156	162	response_format: Annotated[ResponseFormat, Form()] = config.default_response_format,
157	163	temperature: Annotated[float, Form()] = 0.0,
158	164	stream: Annotated[bool, Form()] = False,
	165	+) -> (
	166	+ str
	167	+ \| TranscriptionJsonResponse
	168	+ \| TranscriptionVerboseJsonResponse
	169	+ \| StreamingResponse
159	170	):
160	171	start = time.perf_counter()
161	172	whisper = load_model(model)
...	...	@@ -201,7 +212,10 @@
201	212
202	213	# https://platform.openai.com/docs/api-reference/audio/createTranscription
203	214	# https://github.com/openai/openai-openapi/blob/master/openapi.yaml#L8915
204		-@app.post("/v1/audio/transcriptions")
	215	+@app.post(
	216	+ "/v1/audio/transcriptions",
	217	+ response_model=str \| TranscriptionJsonResponse \| TranscriptionVerboseJsonResponse,
	218	+)
205	219	def transcribe_file(
206	220	file: Annotated[UploadFile, Form()],
207	221	model: Annotated[ModelName, Form()] = config.whisper.model,
...	...	@@ -214,6 +228,11 @@
214	228	Form(alias="timestamp_granularities[]"),
215	229	] = ["segment"],
216	230	stream: Annotated[bool, Form()] = False,
	231	+) -> (
	232	+ str
	233	+ \| TranscriptionJsonResponse
	234	+ \| TranscriptionVerboseJsonResponse
	235	+ \| StreamingResponse
217	236	):
218	237	start = time.perf_counter()
219	238	whisper = load_model(model)

...	...	@@ -130,8 +130,6 @@
130	130
131	131
132	132	class ModelObject(BaseModel):
133		- model_config = ConfigDict(populate_by_name=True)
134		-
135	133	id: str
136	134	"""The model identifier, which can be referenced in the API endpoints."""
137	135	created: int
...	...	@@ -140,3 +138,29 @@
140	138	"""The object type, which is always "model"."""
141	139	owned_by: str
142	140	"""The organization that owns the model."""
	141	+
	142	+ model_config = ConfigDict(
	143	+ populate_by_name=True,
	144	+ json_schema_extra={
	145	+ "examples": [
	146	+ {
	147	+ "id": "Systran/faster-whisper-large-v3",
	148	+ "created": 1700732060,
	149	+ "object": "model",
	150	+ "owned_by": "Systran",
	151	+ },
	152	+ {
	153	+ "id": "Systran/faster-distil-whisper-large-v3",
	154	+ "created": 1711378296,
	155	+ "object": "model",
	156	+ "owned_by": "Systran",
	157	+ },
	158	+ {
	159	+ "id": "bofenghuang/whisper-large-v2-cv11-french-ct2",
	160	+ "created": 1687968011,
	161	+ "object": "model",
	162	+ "owned_by": "bofenghuang",
	163	+ },
	164	+ ]
	165	+ },
	166	+ )

Delete comment