Anuj Solanki

GSoC Week 04 Summary

Date: 17 June 2024 - 23 June 2024

Introduction

In the fourth week of my Google Summer of Code (GSoC) project, I focused on integrating Whisper AI for online mode. The primary objectives were to resolve existing errors in python3-scipy and other recipes, refine the meta-whisper layer recipes, and containerize the Whisper AI service.

Tasks I Have Completed This Week

Resolved Errors in agl-service-voiceagent:
- Resolved issues related in python3-scipy and python3-scikit-learn recipes.
- Fixed errors in the voice-agent flutter app recipe.
Refactored Recipes in meta-whisper Layer:
- Added licenses and Removed unnecessary recipes from the meta-whisper layer.
Containerized Whisper AI Using Docker:
- Created a Docker container for Whisper ASR service.
- Implemented a gRPC-based connection between the containerized Whisper service and a Python client.
- Whisper Container Github Repo

Use agl-whisper-docker:

Clone the Repository:

git clone https://github.com/Anuj-S62/agl-whisper-docker

Build the Docker Image:
```
docker build -t whisper-asr-service .
```

Run the Docker Container:

docker run -p 50051:50051 whisper-asr-service

Access the Whisper Service:
- The Whisper service will be available at localhost:50051.
- Use the Python client to interact with the service.
```
python3 python3 agl_whisper_client_example.py --ip 127.0.0.1 --port 50051 --audio_path audio_file_path
```

Tasks for Next Week

Integrate the containerized Whisper AI into the agl-service-voiceagent.
Test Whisper AI with agl-service-voiceagent on an AGL image.

Conclusion

This week was highly productive, with significant progress made in error resolution, recipe refinement, and the containerization of Whisper AI. I am looking forward to integrating Whisper AI into the AGL service voice agent and validating its functionality in the AGL environment.