A-I(R-A-G) Code Search

• Embedding model use the cohere's embedding models, using AWS Bedrock boto3 services
• ML algorthims used open-sourced faiss's squared Euclidean (L2) distance
• Socket & Multi-threats applied, port 3000, 8000 used, auto socket-based port verification, and subprocess handling with proper cleanup
• Auto Logging('.logs/results.json')
• Cross-OS is available to Windows, MacOS, Linux, Windows(WSL)

!Note:

1. For Similarity Search counting the distance, use Euclidean (L2) distance is better. Don't mixed up with cosine similarity. Two differnt concepts
   
2. With more data, performance of the code search enginee is better. e.g. 10k+ items (take reference with the data's attribute example in ./data
3. TroubleShoot, if the FastAPI server fails to start through run.py, try running it directly:

# Ensure you are in your virtual environment first
# Windows
venv\Scripts\activate
# macOS/Linux/WSL
source venv/bin/activate

# Ensure you have `pip install -r requirements`

# Then run the FastAPI server directly
python -m src.api.main

In the demo example, it only use 10 rows(just for demo purpose), thus the performance bad is normal.