Project push

.gitignore

@@ -0,0 +1,5 @@
+# Virtual Env
+.venv
+
+# Local data
+.DS_Store

README.md

@@ -0,0 +1,78 @@
+# Audio Summary with local LLM
+
+This tool is designed to provide a quick and concise summary of audio and video files. It supports summarizing content either from a local file or directly from YouTube. The tool uses Whisper for transcription and a local version of Mistral AI (Ollama) for generating summaries.
+
+> [!TIP]  
+> It is possible to change the model you wish to use.
+> To do this, change the `OLLAMA_MODEL` variable, and download the associated model via [ollama](https://github.com/ollama/ollama)
+
+## Features
+
+- **YouTube Integration**: Download and summarize content directly from YouTube.
+- **Local File Support**: Summarize audio files available on your local disk.
+- **Transcription**: Converts audio content to text using Whisper.
+- **Summarization**: Generates a concise summary using Mistral AI (Ollama).
+
+## Prerequisites
+
+Before you start using this tool, you need to install the following dependencies:
+
+- Python 3.8 or higher
+- `pytube` for downloading videos from YouTube.
+- `pathlib`for local file
+- `openai-whisper` for audio transcription.
+- [Ollama](https://ollama.com) for LLM model management.
+- `ffmpeg` (required for whisper)
+
+## Installation
+
+### Python Requirements
+
+Clone the repository and install the required Python packages:
+
+```bash
+git clone https://github.com/damienarnodo/audio-summary-with-local-LLM.git
+cd audio-summary-with-local-LLM
+pip install -r src/requirements.txt
+```
+
+### LLM Requierement
+
+[Download and install](https://ollama.com) Ollama to carry out LLM Management
+More details about LLM model supported can be discribe on the Ollama [github](https://github.com/ollama/ollama).
+
+Download and use Mistral model :
+
+```bash
+ollama pull mistral
+
+## Test the access :
+ollama run mistral "tell me a joke"
+```
+
+## Usage
+
+The tool can be executed with the following command line options:
+
+- `--from-youtube`: To download and summarize a video from YouTube.
+- `--from-local`: To load and summarize an audio or video file from the local disk.
+
+### Examples
+
+1. **Summarizing a YouTube video:**
+
+   ```bash
+   python src/summary.py --from-youtube <YouTube-Video-URL>
+   ```
+
+2. **Summarizing a local audio file:**
+
+   ```bash
+   python src/summary.py --from-local <path-to-audio-file>
+   ```
+
+The output summary will be saved in a markdown file in the specified output directory.
+
+## Output
+
+The summarized content is saved as a markdown file named `summary.md` in the current working directory. This file includes the transcribed text and its corresponding summary.

src/requirements.txt

@@ -0,0 +1,3 @@
+openai-whisper==20231117
+pytube==15.0.0
+ollama==0.1.8

src/summary.py

@@ -0,0 +1,98 @@
+import whisper
+import ollama
+import argparse
+from pytube import YouTube
+from pathlib import Path
+
+
+WHISPER_MODEL = "base"
+OLLAMA_MODEL = "mistral"
+
+# Function to download a video from YouTube
+def download_from_youtube(url: str, path: str):
+    yt = YouTube(url)
+    # Filter streams to get the highest resolution progressive mp4 stream
+    stream = yt.streams.filter(file_extension="mp4", only_audio=True).first()
+    # Download the video to the specified path
+    stream.download(Path(path), filename="to_transcribe.mp4")
+
+# Function to transcribe an audio file using the Whisper model
+def transcribe_file(file_path: str, output_file: str) -> str:
+    # Load the Whisper model
+    model = whisper.load_model(WHISPER_MODEL)
+    # Transcribe the audio file
+    transcribe = model.transcribe(file_path)
+    # Save the transcribed text to the specified temporary file
+    with open(output_file, 'w') as tmp_file:
+        tmp_file.write(transcribe["text"])
+        print(f"Transcription saved to file: {output_file}")
+    # Return the transcribed text
+    return transcribe["text"]
+
+# Function to summarize a text using the Ollama model
+def summarize_text(text: str, output_path: str) -> str:
+    # Define the system prompt for the Ollama model
+    system_prompt = f"I would like for you to assume the role of a Technical Expert"
+    # Define the user prompt for the Ollama model
+    user_prompt = f"""Generate a concise summary of the text below.
+    Text : {text}
+    Add a title to the summary.
+    Make sure your summary has useful and true information about the main points of the topic.
+    Begin with a short introduction explaining the topic. If you can, use bullet points to list important details,
+    and finish your summary with a concluding sentence."""
+
+    # Use the Ollama model to generate a summary
+    response = ollama.chat(
+        model=OLLAMA_MODEL,
+        messages=[
+            {
+                "role": "system",
+                "content": system_prompt,
+            },
+            {
+                "role": "user",
+                "content": user_prompt,
+            },
+        ],
+    )
+    # Print the generated summary
+    return response["message"]["content"]
+
+def main():
+    # Parse command line arguments
+    parser = argparse.ArgumentParser(description="Download, transcribe, and summarize audio or video files.")
+    group = parser.add_mutually_exclusive_group(required=True)
+    group.add_argument("--from-youtube", type=str, help="YouTube URL to download.")
+    group.add_argument("--from-local", type=str, help="Path to the local audio file.")
+    parser.add_argument("--output", type=str, default="./summary.md", help="Output markdown file path.")
+    
+    args = parser.parse_args()
+    
+    # Set up data directory
+    data_directory = Path("tmp")
+  
+    if args.from_youtube:
+        # Download from YouTube
+        print(f"Downloading YouTube video from {args.from_youtube}")
+        download_from_youtube(args.from_youtube, str(data_directory))
+        file_path = data_directory / "to_transcribe.mp4"
+    elif args.from_local:
+        # Use local file
+        file_path = args.from_local
+    
+    print(f"Transcribing file: {file_path}")
+    # Transcribe the audio file
+    transcript = transcribe_file(str(file_path), data_directory / "transcript.txt")
+    
+    print("Generating summary...")
+    # Generate summary
+    summary = summarize_text(transcript, "./")
+    
+    # Write summary to a markdown file
+    with open(args.output, "w") as md_file:
+        md_file.write("# Summary\n\n")
+        md_file.write(summary)
+        print(f"Summary written to {args.output}")
+
+if __name__ == "__main__":
+    main()