change structure of the project

Changes:
   Move some modules for it's own file and move config to the `config.py`

app.py

@@ -1,79 +1,14 @@
 import gradio as gr
-from faster_whisper import WhisperModel
 from pathlib import Path
-from dotenv import load_dotenv
-from typing import Optional
-import os
 
+# Подгрузка сервисов
 from services.llm import Llm
+from services.fasterWhisper import FasterWhisper
 
-load_dotenv()
+# Загрузка параметров конфигурации
-
+from config import *
-# Стандартный API ключ
-DEFAULT_API_KEY=os.getenv('API_KEY')
-# Задаем выходную директорию
-OUTPUT_PATH='outputs'
-# Стандартный системный промпт
-DEFAULT_SYSTEM_PROMPT='''You are a diligent university student who has recorded a lecture as an audio file and later transcribed it into raw text. 
-Your task is to rewrite this unstructured transcript into a clear, logically organized, and detailed lecture summary (lecture notes).  
-
-Guidelines:
-1. Structure:
-   - Organize the text into a hierarchy of sections and subsections.  
-   - Use headings, bullet points, or numbering where appropriate.  
-   - Present the material in a logical flow (from introduction → main points → details → examples → conclusion).  
-
-2. Clarity & Cohesion:
-   - Remove filler words, repetitions, and irrelevant fragments.  
-   - Rewrite incomplete sentences into full, grammatically correct sentences.  
-   - Ensure smooth transitions between topics, making the summary feel continuous and well-connected.  
-
-3. Depth & Detail: 
-   - Capture all important concepts, definitions, examples, and explanations from the lecture.  
-   - Expand shorthand or fragmented thoughts into full, precise explanations.  
-   - Where appropriate, rephrase or clarify confusing passages for better understanding.  
-
-4. Accuracy:
-   - Preserve the lecturer’s original meaning, intent, and terminology.  
-   - Avoid adding personal opinions or new information that was not in the lecture.  
-
-5. Style:
-   - Write in a formal, academic tone suitable for study notes.  
-   - Aim for readability: concise sentences, but thorough coverage of concepts.  
-   - Use emphasis (e.g., bold or italic text) only when it improves comprehension.  
-
-Final Output: A cohesive, detailed, and well-structured lecture summary, suitable for later studying and revision.
-Use only russian language!
-'''
-
-faster_whisper_models_list = ['tiny', 'base', 'small', 'medium', 'large-v1', 'large-v2', 'large-v3', 'large', 'distil-large-v2', 'distil-large-v3', 'distil-large-v3.5', 'large-v3-turbo', 'turbo']
-llm_models_list = ['openai/gpt-oss-120b', 'Qwen/Qwen3-235B-A22B-Thinking-2507', 'deepseek-ai/DeepSeek-R1-0528', 'meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8', 'openai/gpt-oss-20b', 'Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar', 'meta-llama/Llama-3.2-90B-Vision-Instruct', 'mistralai/Mistral-Nemo-Instruct-2407', 'Qwen/Qwen2.5-VL-32B-Instruct', 'meta-llama/Llama-3.3-70B-Instruct', 'mistralai/Devstral-Small-2505', 'mistralai/Magistral-Small-2506', 'mistralai/Mistral-Large-Instruct-2411', 'CohereForAI/aya-expanse-32b']
 
 # Функция транскрибации
-def recognize(model, audioFile, beamSize, vadFilter, minSilenceDurationMs, speechPadMs, temp0, temp1, temp2, wordTimestamps, noSpeechThreshold, conditionOnPreviousText):
-    model = WhisperModel(model, device='cuda', compute_type='float16') # Задаем модель
-
-    segments, _ = model.transcribe( # Распознаем текст
-        audioFile,
-        beam_size=beamSize,
-        vad_filter=vadFilter,
-        vad_parameters={
-            "min_silence_duration_ms": minSilenceDurationMs,
-            "speech_pad_ms": speechPadMs
-        },
-        temperature= [temp0, temp1, temp2],
-        word_timestamps=wordTimestamps,
-        no_speech_threshold=noSpeechThreshold,
-        condition_on_previous_text=conditionOnPreviousText
-    )
-
-    text = ''
-
-    for seg in segments:
-        text += f"[{format_timestamp(seg.start)} -> {format_timestamp(seg.end)}] {seg.text}" + '\n'
-    
-    return(text)
-
 def generateByCondition(api_key, llm_model, system_prompt, recognized_text, llm_temperature, is_pipeline_enabled, trigger, isSaveFile):
     llm = Llm(api_key)
 
@@ -98,13 +33,6 @@ def generateByCondition(api_key, llm_model, system_prompt, recognized_text, llm_
     # если нет чекбокса и было событие change
     return gr.skip(), gr.skip()
 
-def format_timestamp(seconds: float) -> str:
-    millis = int(seconds * 1000)
-    hours = millis // (3600 * 1000)
-    minutes = (millis % (3600 * 1000)) // (60 * 1000)
-    seconds_int = (millis % (60 * 1000)) // 1000
-    millis = millis % 1000
-    return f"{hours:02d}:{minutes:02d}:{seconds_int:02d},{millis:03d}"
 
 # Функция сохранения файла
 def saveFile(filename, text):
@@ -123,7 +51,6 @@ def updateButton(isChecked):
     return gr.update(interactive=not isChecked, variant=variant)
 
 
-
 def updateTextbox(isChecked):
     return gr.update(visible=isChecked)
 
@@ -161,7 +88,7 @@ with gr.Blocks() as demo:
                         with gr.Row():
                             # Левая колонка в акордионе
                             with gr.Column():
-                                fastWhisperModel = gr.Dropdown(label='Model', choices=faster_whisper_models_list, interactive=True)
+                                fastWhisperModel = gr.Dropdown(label='Model', choices=FAST_WHISPER_MODELS, value=FAST_WHISPER_MODELS[11], interactive=True)
 
                                 beamSize = gr.Number(label='beam_size', value=8, interactive=True)
                                 noSpeechThreshold  = gr.Number(label='no_speech_threshold', value=0.5, interactive=True)
@@ -188,7 +115,7 @@ with gr.Blocks() as demo:
                             systemPrompt = gr.Textbox(label='', value=DEFAULT_SYSTEM_PROMPT, interactive=True)
 
                         with gr.Row():
-                            llmModel = gr.Dropdown(label='models', choices=llm_models_list, value=llm_models_list[0], interactive=True)
+                            llmModel = gr.Dropdown(label='models', choices=LLM_MODELS, value=LLM_MODELS[0], interactive=True)
                             llmTemperature = gr.Number(label='Temperature', value=0.8, interactive=True )
 
         with gr.Column():
@@ -206,7 +133,7 @@ with gr.Blocks() as demo:
     isPipelineEnabledCheckbox.change(updateButton, inputs=[isPipelineEnabledCheckbox], outputs=refineTextBtn)
     saveFileCheckbox.change(updateTextbox, inputs=saveFileCheckbox, outputs=filename)
 
-    recognizeBtn.click(recognize, outputs=[recognizedText], inputs=[fastWhisperModel, audioFile, beamSize, vadFilter, minSilenceDurationMs, speechPadMs, temp0, temp1, temp2, wordTimestamps, noSpeechThreshold, conditionOnPreviousText])
+    recognizeBtn.click(FasterWhisper.recognize, outputs=[recognizedText], inputs=[fastWhisperModel, audioFile, beamSize, vadFilter, minSilenceDurationMs, speechPadMs, temp0, temp1, temp2, wordTimestamps, noSpeechThreshold, conditionOnPreviousText])
     
     # Если пайплайн включен то тогда делаем автоматически
     # автоматический пайплайн
@@ -223,5 +150,4 @@ with gr.Blocks() as demo:
         outputs=[refinedText, refinedTextMD]
     )
 
-
 demo.launch()

config.py

@@ -0,0 +1,48 @@
+import os
+from dotenv import load_dotenv
+
+load_dotenv()
+
+FAST_WHISPER_MODELS = ['tiny', 'base', 'small', 'medium', 'large-v1', 'large-v2', 'large-v3', 'large', 'distil-large-v2', 'distil-large-v3', 'distil-large-v3.5', 'large-v3-turbo', 'turbo']
+LLM_MODELS = ['openai/gpt-oss-120b', 'Qwen/Qwen3-235B-A22B-Thinking-2507', 'deepseek-ai/DeepSeek-R1-0528', 'meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8', 'openai/gpt-oss-20b', 'Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar', 'meta-llama/Llama-3.2-90B-Vision-Instruct', 'mistralai/Mistral-Nemo-Instruct-2407', 'Qwen/Qwen2.5-VL-32B-Instruct', 'meta-llama/Llama-3.3-70B-Instruct', 'mistralai/Devstral-Small-2505', 'mistralai/Magistral-Small-2506', 'mistralai/Mistral-Large-Instruct-2411', 'CohereForAI/aya-expanse-32b']
+
+# Стандартный API ключ
+DEFAULT_API_KEY=os.getenv('API_KEY')
+# Задаем выходную директорию
+OUTPUT_PATH='outputs'
+
+# Стандартный системный промпт
+OUTPUT_PATH='outputs'
+
+DEFAULT_SYSTEM_PROMPT='''You are a diligent university student who has recorded a lecture as an audio file and later transcribed it into raw text. 
+Your task is to rewrite this unstructured transcript into a clear, logically organized, and detailed lecture summary (lecture notes).  
+
+Guidelines:
+1. Structure:
+   - Organize the text into a hierarchy of sections and subsections.  
+   - Use headings, bullet points, or numbering where appropriate.  
+   - Present the material in a logical flow (from introduction → main points → details → examples → conclusion).  
+
+2. Clarity & Cohesion:
+   - Remove filler words, repetitions, and irrelevant fragments.  
+   - Rewrite incomplete sentences into full, grammatically correct sentences.  
+   - Ensure smooth transitions between topics, making the summary feel continuous and well-connected.  
+
+3. Depth & Detail: 
+   - Capture all important concepts, definitions, examples, and explanations from the lecture.  
+   - Expand shorthand or fragmented thoughts into full, precise explanations.  
+   - Where appropriate, rephrase or clarify confusing passages for better understanding.  
+
+4. Accuracy:
+   - Preserve the lecturer’s original meaning, intent, and terminology.  
+   - Avoid adding personal opinions or new information that was not in the lecture.  
+
+5. Style:
+   - Write in a formal, academic tone suitable for study notes.  
+   - Aim for readability: concise sentences, but thorough coverage of concepts.  
+   - Use emphasis (e.g., bold or italic text) only when it improves comprehension.  
+
+Final Output: A cohesive, detailed, and well-structured lecture summary, suitable for later studying and revision.
+Use only russian language!
+'''
+

services/fasterWhisper.py

@@ -0,0 +1,39 @@
+from faster_whisper import WhisperModel
+
+class FasterWhisper:
+    def __init__(self):
+        pass
+
+    def recognize(self, model, audioFile, beamSize, vadFilter, minSilenceDurationMs, speechPadMs, temp0, temp1, temp2, wordTimestamps, noSpeechThreshold, conditionOnPreviousText):
+        model = WhisperModel(model, device='cuda', compute_type='float16') # Задаем модель
+
+        segments, _ = model.transcribe( # Распознаем текст
+            audioFile,
+            beam_size=beamSize,
+            vad_filter=vadFilter,
+            vad_parameters={
+                "min_silence_duration_ms": minSilenceDurationMs,
+                "speech_pad_ms": speechPadMs
+            },
+            temperature= [temp0, temp1, temp2],
+            word_timestamps=wordTimestamps,
+            no_speech_threshold=noSpeechThreshold,
+            condition_on_previous_text=conditionOnPreviousText
+        )
+
+        text = ''
+
+        for seg in segments:
+            text += f"[{self.format_timestamp(seg.start)} -> {self.format_timestamp(seg.end)}] {seg.text}" + '\n'
+        
+        return(text)
+
+    def format_timestamp(self, seconds: float) -> str:
+        millis = int(seconds * 1000)
+        hours = millis // (3600 * 1000)
+        minutes = (millis % (3600 * 1000)) // (60 * 1000)
+        seconds_int = (millis % (60 * 1000)) // 1000
+        millis = millis % 1000
+        return f"{hours:02d}:{minutes:02d}:{seconds_int:02d},{millis:03d}"
+
+