Support OpenAI Whisper API Integration for Enhanced Transcription Flexibility

README.md

@@ -16,16 +16,19 @@ GroupLang-secretary-bot is a Telegram bot that transcribes voice messages, summa
 
 ## Features
 
-- Transcribes voice messages using AWS Transcribe
+- Transcribes voice messages using either AWS Transcribe or OpenAI Whisper API
 - Summarizes transcribed text using a custom API
 - Allows users to tip for the service
 - Secures handling of API keys and tokens
 - Is deployable as an AWS Lambda function
+- Configurable transcription service selection
 
 ## Prerequisites
 
 - Poetry for dependency management
-- AWS account with Transcribe access
+- One of the following transcription services:
+  - AWS account with Transcribe access
+  - OpenAI API key with Whisper API access
 - Telegram Bot Token
 - MarketRouter API Key
 
@@ -70,13 +73,25 @@ To quickly get started with the GroupLang-secretary-bot, follow these steps:
 
 1. Set up environment variables:
    - `TELEGRAM_BOT_TOKEN`: Your Telegram Bot Token
+   - `MARKETROUTER_API_KEY`: Your MarketRouter API Key
+   - `TRANSCRIPTION_SERVICE`: Choose between 'aws' or 'openai' (default: 'aws')
+
+   For AWS Transcribe:
    - `AWS_ACCESS_KEY_ID`: Your AWS Access Key ID
    - `AWS_SECRET_ACCESS_KEY`: Your AWS Secret Access Key
-   - `MARKETROUTER_API_KEY`: Your MarketRouter API Key
-
-2. Configure AWS credentials:
-   - Either set up the AWS CLI with `aws configure` or use environment variables as mentioned above.
-   - Ensure that your AWS IAM user has the necessary permissions for AWS Transcribe.
+   - `AWS_REGION`: AWS region (default: 'us-east-1')
+
+   For OpenAI Whisper:
+   - `OPENAI_API_KEY`: Your OpenAI API Key
+
+2. Configure credentials based on your chosen transcription service:
+
+   For AWS Transcribe:
+   - Either set up the AWS CLI with `aws configure` or use environment variables as mentioned above
+   - Ensure that your AWS IAM user has the necessary permissions for AWS Transcribe
+
+   For OpenAI Whisper:
+   - Ensure your OpenAI API key has access to the Whisper API
 
 1. Activate the Poetry virtual environment:
    ```
@@ -139,7 +154,12 @@ poetry update package_name
 
 The bot uses the following external APIs:
 
-- AWS Transcribe: For audio transcription
+- For audio transcription (configurable):
+  - AWS Transcribe: Amazon's speech-to-text service
+  - OpenAI Whisper API: OpenAI's speech-to-text service
 - MarketRouter API: For text summarization and reward submission
 
-Refer to the respective documentation for more details on these APIs.
+Refer to the respective documentation for more details:
+- [AWS Transcribe Documentation](https://docs.aws.amazon.com/transcribe/)
+- [OpenAI Whisper API Documentation](https://platform.openai.com/docs/guides/speech-to-text)
+- MarketRouter API Documentation

bot_handlers.py

@@ -1,16 +1,16 @@
 import logging
 import os
 from typing import Dict, Any
-from services import AWSServices, AudioTranscriber, TextSummarizer
+from services import TranscriberFactory, TextSummarizer
 from utils.telegram_utils import send_message, get_telegram_file_url
 from utils.message_utils import format_response, create_tip_button
+from config import Config
 
 logger = logging.getLogger(__name__)
 
 # Initialize services
-aws_services = AWSServices()
-audio_transcriber = AudioTranscriber(aws_services)
-text_summarizer = TextSummarizer(os.environ.get('MARKETROUTER_API_KEY'))
+audio_transcriber = TranscriberFactory.create_transcriber(Config)
+text_summarizer = TextSummarizer(Config.MARKETROUTER_API_KEY)
 
 def handle_update(update: Dict[str, Any]) -> None:
     if 'message' in update:

config.py

@@ -6,3 +6,5 @@ class Config:
     MARKETROUTER_API_KEY = os.environ.get('MARKETROUTER_API_KEY')
     AWS_ACCESS_KEY_ID = os.environ.get('AWS_ACCESS_KEY_ID')
     AWS_SECRET_ACCESS_KEY = os.environ.get('AWS_SECRET_ACCESS_KEY')
+    OPENAI_API_KEY = os.environ.get('OPENAI_API_KEY')
+    TRANSCRIPTION_SERVICE = os.environ.get('TRANSCRIPTION_SERVICE', 'aws')  # 'aws' or 'openai'

services.py

@@ -4,8 +4,10 @@
 import time
 import uuid
 import logging
+import os
 from io import BytesIO
 from botocore.exceptions import ClientError
+from abc import ABC, abstractmethod
 
 logger = logging.getLogger(__name__)
 
@@ -50,7 +52,19 @@ def start_transcription_job(self, job_name, media_uri, media_format='ogg', langu
     def get_transcription_job_status(self, job_name):
         return self.transcribe_client.get_transcription_job(TranscriptionJobName=job_name)
 
-class AudioTranscriber:
+from abc import ABC, abstractmethod
+
+class BaseTranscriber(ABC):
+    @abstractmethod
+    def transcribe_audio(self, file_url: str) -> str:
+        pass
+
+    def _download_audio(self, file_url: str) -> bytes:
+        response = requests.get(file_url)
+        response.raise_for_status()
+        return response.content
+
+class AWSTranscriber(BaseTranscriber):
     def __init__(self, aws_services: AWSServices):
         self.aws_services = aws_services
         self.bucket_name = 'audio-transcribe-temp'
@@ -77,11 +91,6 @@ def transcribe_audio(self, file_url: str) -> str:
             logger.error(f"An error occurred: {e}")
             raise
 
-    def _download_audio(self, file_url: str) -> bytes:
-        response = requests.get(file_url)
-        response.raise_for_status()
-        return response.content
-
     def _wait_for_transcription(self, job_name: str) -> str:
         while True:
             status = self.aws_services.get_transcription_job_status(job_name)
@@ -95,6 +104,68 @@ def _wait_for_transcription(self, job_name: str) -> str:
         else:
             raise Exception("Transcription failed")
 
+class OpenAITranscriber(BaseTranscriber):
+    def __init__(self, api_key: str):
+        self.api_key = api_key
+        self.api_url = "https://api.openai.com/v1/audio/transcriptions"
+
+    def transcribe_audio(self, file_url: str) -> str:
+        try:
+            audio_content = self._download_audio(file_url)
+
+            # Save audio content to a temporary file
+            temp_file = f'/tmp/audio_{uuid.uuid4()}.ogg'
+            with open(temp_file, 'wb') as f:
+                f.write(audio_content)
+
+            # Prepare the request
+            headers = {
+                "Authorization": f"Bearer {self.api_key}"
+            }
+
+            files = {
+                'file': ('audio.ogg', open(temp_file, 'rb'), 'audio/ogg'),
+                'model': (None, 'whisper-1'),
+            }
+
+            # Make the request to OpenAI's Whisper API
+            response = requests.post(
+                self.api_url,
+                headers=headers,
+                files=files
+            )
+            response.raise_for_status()
+
+            # Clean up the temporary file
+            os.remove(temp_file)
+
+            result = response.json()
+            return result.get('text', '')
+
+        except Exception as e:
+            logger.error(f"An error occurred with OpenAI transcription: {e}")
+            raise
+        finally:
+            # Ensure temp file is removed even if an error occurs
+            if 'temp_file' in locals() and os.path.exists(temp_file):
+                os.remove(temp_file)
+
+class TranscriberFactory:
+    @staticmethod
+    def create_transcriber(config) -> BaseTranscriber:
+        service_type = config.TRANSCRIPTION_SERVICE.lower()
+
+        if service_type == 'aws':
+            aws_services = AWSServices(region_name=config.AWS_REGION)
+            return AWSTranscriber(aws_services)
+        elif service_type == 'openai':
+            if not config.OPENAI_API_KEY:
+                raise ValueError("OpenAI API key is required for OpenAI transcription service")
+            return OpenAITranscriber(config.OPENAI_API_KEY)
+        else:
+            raise ValueError(f"Unsupported transcription service: {service_type}")
+
+
 class TextSummarizer:
     def __init__(self, api_key: str):
         self.api_key = api_key