Add Support for OpenAI Whisper API as Additional Transcription Backend

README.md

@@ -16,7 +16,8 @@ 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
+- Flexible choice of transcription service based on configuration
 - Summarizes transcribed text using a custom API
 - Allows users to tip for the service
 - Secures handling of API keys and tokens
@@ -25,7 +26,9 @@ GroupLang-secretary-bot is a Telegram bot that transcribes voice messages, summa
 ## Prerequisites
 
 - Poetry for dependency management
-- AWS account with Transcribe access
+- One of the following transcription services:
+  - AWS account with Transcribe access, OR
+  - OpenAI API key for Whisper API access
 - Telegram Bot Token
 - MarketRouter API Key
 
@@ -70,13 +73,18 @@ 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' (default) or 'openai'
+
+2. Configure transcription service credentials:
+   For AWS Transcribe (if using 'aws' service):
    - `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
+   - `AWS_REGION`: AWS region (defaults to 'us-east-1')
+   - Ensure that your AWS IAM user has the necessary permissions for AWS Transcribe
 
-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.
+   For OpenAI Whisper (if using 'openai' service):
+   - `OPENAI_API_KEY`: Your OpenAI API key with access to the Whisper API
 
 1. Activate the Poetry virtual environment:
    ```
@@ -139,7 +147,12 @@ 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 recognition model
 - 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 on these APIs:
+- [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 (refer to your API provider)

bot_handlers.py

@@ -1,16 +1,30 @@
 import logging
 import os
 from typing import Dict, Any
-from services import AWSServices, AudioTranscriber, TextSummarizer
+from services import (
+    AWSServices, AWSTranscriber, OpenAITranscriber, 
+    TextSummarizer, BaseTranscriber
+)
 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'))
+def get_transcriber() -> BaseTranscriber:
+    if Config.TRANSCRIPTION_SERVICE == 'openai':
+        if not Config.OPENAI_API_KEY:
+            raise ValueError("OpenAI API key is required when using OpenAI transcription service")
+        return OpenAITranscriber(Config.OPENAI_API_KEY)
+    else:  # default to AWS
+        if not (Config.AWS_ACCESS_KEY_ID and Config.AWS_SECRET_ACCESS_KEY):
+            raise ValueError("AWS credentials are required when using AWS transcription service")
+        aws_services = AWSServices(Config.AWS_REGION)
+        return AWSTranscriber(aws_services)
+
+audio_transcriber = get_transcriber()
+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'

pyproject.toml

@@ -17,6 +17,7 @@ requests = "^2.32.3"
 nltk = "^3.9.1"
 langdetect = "^1.0.9"
 mangum = "^0.18.0"
+openai = "^1.12.0"
 
 [tool.poetry.dev-dependencies]
 # Add any development dependencies here

services.py

@@ -50,7 +50,20 @@ 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
+import openai
+
+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 +90,6 @@ 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 +103,34 @@ def _wait_for_transcription(self, job_name: str) -> str:
         else:
             raise Exception("Transcription failed")
 
+class OpenAITranscriber(BaseTranscriber):
+    def __init__(self, api_key: str):
+        self.client = openai.OpenAI(api_key=api_key)
+
+    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)
+
+            # Transcribe using OpenAI Whisper API
+            with open(temp_file, 'rb') as audio_file:
+                transcription = self.client.audio.transcriptions.create(
+                    model="whisper-1",
+                    file=audio_file,
+                    response_format="text"
+                )
+
+            # Clean up temporary file
+            os.remove(temp_file)
+            return transcription
+        except Exception as e:
+            logger.error(f"An error occurred during OpenAI transcription: {e}")
+            raise
+
 class TextSummarizer:
     def __init__(self, api_key: str):
         self.api_key = api_key