Add MP3 transcription support

.dockerignore

@@ -0,0 +1,16 @@
+# Git
+.git
+.gitignore
+
+# Python
+.python-version
+__pycache__/
+*.py[cod]
+.venv/
+.env
+
+# Docker
+Dockerfile
+
+# Other
+.ropeproject

.github/workflows/docker-build.yml

@@ -0,0 +1,25 @@
+name: Docker Build CI
+
+on:
+  push:
+    branches-ignore:
+      - 'main'
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+
+    - name: Set up Docker Buildx
+      uses: docker/setup-buildx-action@v3
+
+    - name: Build Docker image
+      uses: docker/build-push-action@v5
+      with:
+        context: .
+        push: false
+        load: true
+        tags: agent-utils:build-test

.github/workflows/docker-publish.yml

@@ -0,0 +1,30 @@
+name: Docker Image CI
+
+on:
+  push:
+    branches: [ "main" ]
+
+jobs:
+  build-and-publish:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v4
+
+    - name: Log in to the GitHub Container Registry
+      uses: docker/login-action@v3
+      with:
+        registry: ghcr.io
+        username: ${{ github.actor }}
+        password: ${{ secrets.GITHUB_TOKEN }}
+
+    - name: Build and push Docker image
+      uses: docker/build-push-action@v5
+      with:
+        context: .
+        push: true
+        tags: ghcr.io/${{ github.repository }}:latest

.python-version

@@ -0,0 +1 @@
+3.10

Dockerfile

@@ -0,0 +1,30 @@
+# Use an official Python runtime as a parent image
+FROM python:3.10-slim
+
+# Set environment variables
+ENV PYTHONDONTWRITEBYTECODE=1
+ENV PYTHONUNBUFFERED=1
+
+# Install system dependencies (ffmpeg for pydub)
+RUN apt-get update && apt-get install -y ffmpeg && rm -rf /var/lib/apt/lists/*
+
+# Set the working directory in the container
+WORKDIR /app
+
+# Install uv
+RUN pip install uv
+
+# Copy the dependency files
+COPY pyproject.toml uv.lock ./
+
+# Install dependencies using uv
+RUN uv export --no-dev | uv pip install --system -r -
+
+# Copy the application code
+COPY ./app ./app
+
+# Expose the port the app runs on
+EXPOSE 8000
+
+# Run the application
+CMD ["fastapi", "run", "--host", "0.0.0.0", "--port", "8000"]

README.md

@@ -1,2 +1,35 @@
 # agent-utils
+
 Various utility APIs for agents
+
+## Usage
+
+Run directly:
+
+```sh
+uv run fastapi [dev|run]
+```
+
+Run the API in development or poduction mode.
+
+Run production mode in a container:
+
+```sh
+docker run -p 8000:8000 --env-file .env agent-utils
+```
+
+## Functions
+
+Find the API documentation at http://…:8000/docs.
+
+### Transcribe audio
+
+Post the URL of an audio file and get a text transcript returned. Currently works with MP3.
+
+## Configuration
+
+The app needs an OpenAI API key to work, in the `.env` file.
+
+```toml
+OPENAI_API_KEY=… # Open AI, API key for access to the Whisper API
+```

app/__init__.py

@@ -0,0 +1 @@
+from . import mp3transcribe

app/main.py

@@ -0,0 +1,38 @@
+from typing import Union
+
+from fastapi import FastAPI, HTTPException
+from dotenv import load_dotenv
+
+from .mp3transcribe import Mp3Transcribe, AudioFile
+
+load_dotenv()
+app = FastAPI()
+
+
+@app.get("/")
+def read_root():
+    return {"Hello": "World"}
+
+
+
+@app.post("/mp3transcribe/")
+def transcribe_mp3(audio_file: AudioFile):
+    """
+    Transcribes an MP3 audio file.
+
+    This endpoint accepts an MP3 audio file, downloads it, and transcribes the audio content to text.
+
+    Args:
+        audio_file (mp3transcribe.AudioFile): An object containing the URL and language of the MP3 audio file to be transcribed.
+
+    Returns:
+        dict: A dictionary containing the transcribed text.
+
+    Raises:
+        HTTPException: If the audio file cannot be downloaded, an HTTP 500 error is raised.
+    """
+    mp3 = Mp3Transcribe(audio_file.url, audio_file.lang)
+    if not mp3.download():
+        raise HTTPException(status_code=500, detail="Failed to download audio file")
+    transcript = mp3.transcribe()
+    return {"text": transcript}

app/mp3transcribe/__init__.py

@@ -0,0 +1,81 @@
+import logging
+from pathlib import Path
+from tempfile import TemporaryDirectory
+
+from pydantic import BaseModel
+import requests
+from openai import OpenAI
+from pydub import AudioSegment
+
+
+
+class AudioFile(BaseModel):
+    url: str
+    lang: str = 'en'
+
+
+class Mp3Transcribe:
+    """
+    A class to handle downloading and transcribing MP3 audio files using OpenAI's transcription service.
+
+    Attributes:
+        url (str): The URL of the MP3 file to download.
+        lang (str): The language of the audio for transcription. Default is 'en'.
+        client (OpenAI): An instance of the OpenAI client.
+        temp_dir (TemporaryDirectory): A temporary directory to store downloaded and processed files.
+        audio (Path): The path to the downloaded MP3 file.
+        transcript (Path): The path to the generated transcript file.
+
+    Methods:
+        download() -> bool:
+            Downloads the MP3 file from the specified URL and saves it to a temporary directory.
+            Returns True if the download is successful, otherwise False.
+
+        transcribe() -> str:
+            Transcribes the downloaded MP3 file in 10-minute segments using OpenAI's transcription service.
+            Saves the transcript to a text file in the temporary directory.
+            Returns the transcribed text.
+    """
+    def __init__(self, url, lang='en'):
+        self.url = url
+        self.lang = lang
+        self.client = OpenAI()
+        self.temp_dir = TemporaryDirectory()
+
+    def download(self) -> bool:
+        self.audio = Path(self.temp_dir.name).joinpath('audio.mp3')
+        logging.debug(f'Downloading {self.url} to {self.audio}')
+        response = requests.get(self.url)
+        if response.status_code != 200:
+            logging.error(f'Failed to download {self.url}, response {response}')
+            return False
+        with open(self.audio, 'wb') as f:
+            f.write(response.content)
+        logging.debug(f'Downloaded {self.url} to {self.audio}')
+        return True
+
+    def transcribe(self):
+        audio_file = AudioSegment.from_mp3(self.audio)
+        ten_minutes = 10 * 60 * 1000
+        audio_10s = audio_file[::ten_minutes]
+        self.transcript = Path(self.temp_dir.name).joinpath('transcript.txt')
+        with open(self.transcript, 'w') as t:
+            i = 0
+            for audio in audio_10s:
+                logging.debug(f'Transcribing segment {i}')
+                seg10_name = Path(self.temp_dir.name).joinpath(f'd10-{i}.mp3')
+                audio.export(seg10_name, format="mp3")
+                with open(seg10_name, 'rb') as f:
+                    text = self.client.audio.transcriptions.create(
+                        language=self.lang,
+                        file=f,
+                        model="whisper-1",
+                        #prompt=prompt
+                        response_format="text"
+                    )
+                    #prompt = transcript
+                    t.write(text)
+                i += 1 
+            logging.debug(f'Transcribed {self.audio} to {self.transcript}')
+        return self.transcript.read_text()
+

pyproject.toml

@@ -0,0 +1,15 @@
+[project]
+name = "agent-utils"
+version = "0.1.0"
+classifiers = ["Private :: Do Not Upload"]
+description = "Utility APIs for agents"
+readme = "README.md"
+requires-python = ">=3.10"
+dependencies = [
+    "fastapi[standard]>=0.115.5",
+    "openai>=1.55.3",
+    "pydantic-settings>=2.6.1",
+    "pydub>=0.25.1",
+    "python-dotenv>=1.0.1",
+    "requests>=2.32.3",
+]