fix: circular import

Taskfile.yaml

@@ -15,7 +15,7 @@ tasks:
   create-multi-arch-builder: docker buildx create --name main --driver=docker-container
   build-and-push:
     cmds:
-      - docker compose build --builder main
+      - docker compose build --builder main --push
     sources:
       - Dockerfile.*
       - speaches/*.py

speaches/config.py

@@ -3,8 +3,6 @@ import enum
 from pydantic import BaseModel, Field
 from pydantic_settings import BaseSettings, SettingsConfigDict
 
-from speaches.server_models import ResponseFormat
-
 SAMPLES_PER_SECOND = 16000
 BYTES_PER_SAMPLE = 2
 BYTES_PER_SECOND = SAMPLES_PER_SECOND * BYTES_PER_SAMPLE
@@ -12,6 +10,15 @@ BYTES_PER_SECOND = SAMPLES_PER_SECOND * BYTES_PER_SAMPLE
 # 1 SECOND OF AUDIO = 32000 BYTES = 16000 SAMPLES
 
 
+# https://platform.openai.com/docs/api-reference/audio/createTranscription#audio-createtranscription-response_format
+class ResponseFormat(enum.StrEnum):
+    TEXT = "text"
+    JSON = "json"
+    VERBOSE_JSON = "verbose_json"
+    # VTT = "vtt"
+    # SRT = "srt"
+
+
 # https://huggingface.co/Systran
 class Model(enum.StrEnum):
     TINY_EN = "tiny.en"

speaches/main.py

@@ -17,10 +17,11 @@ from faster_whisper.vad import VadOptions, get_speech_timestamps
 from speaches import utils
 from speaches.asr import FasterWhisperASR
 from speaches.audio import AudioStream, audio_samples_from_file
-from speaches.config import SAMPLES_PER_SECOND, Language, Model, config
+from speaches.config import (SAMPLES_PER_SECOND, Language, Model,
+                             ResponseFormat, config)
 from speaches.core import Transcription
 from speaches.logger import logger
-from speaches.server_models import (ResponseFormat, TranscriptionJsonResponse,
+from speaches.server_models import (TranscriptionJsonResponse,
                                     TranscriptionVerboseJsonResponse)
 from speaches.transcriber import audio_transcriber
 

speaches/server_models.py

@@ -1,7 +1,5 @@
 from __future__ import annotations
 
-import enum
-
 from faster_whisper.transcribe import Segment, TranscriptionInfo, Word
 from pydantic import BaseModel
 
@@ -9,15 +7,6 @@ from speaches import utils
 from speaches.core import Transcription
 
 
-# https://platform.openai.com/docs/api-reference/audio/createTranscription#audio-createtranscription-response_format
-class ResponseFormat(enum.StrEnum):
-    TEXT = "text"
-    JSON = "json"
-    VERBOSE_JSON = "verbose_json"
-    # VTT = "vtt"
-    # SRT = "srt"
-
-
 # https://platform.openai.com/docs/api-reference/audio/json-object
 class TranscriptionJsonResponse(BaseModel):
     text: str