CS510 LLM Agents - Assignment 1

Group 6

Team Members & Contribution

• Chase Jamieson
  • Question 2 code implementation
  • Bioasq datasets study
• Chia Wei Hsu
  • Question 1 code implementation
  • Ollama backend implementation
  • Project status tracking
  • Text UI, merged code env maintenance
• Satvik Mudgal
  • Question 3 code implementation
  • Streamlit user interface implementation

Dependencies

• This bundle requires conda.
• Conda environment is listed in the file environment.yml.
• Pip dependencies are listed in the file requirements.txt.
• If you wish to use Ollama as the LM backend, you must install Ollama:
  • curl -fsSL https://ollama.com/install.sh | sh (Linux)
  • brew install ollama (if you are on Mac)
  • then run ollama serve

(Optional) Download pre-packaged embeddings

You can download the embeddings on the releases page to avoid running the embeddings on your machine.

https://github.com/planetaska/rag6/releases/tag/embeddings

After download, simply unzip the file then put the directory ragqa_bge_index and any additional embeddings under embeddings/.

Running the program (TEXT UI)

1. Create a local .env file from the example and add your credentials

cp .env.example .env

# Then update the .env file with your own keys

2. Rebuild the environment:

# rebuild env
conda env create -f environment.yml

# activate
conda activate rag6

# run the program
python main.py

3. (Optional) If you wish to use Ollama locally:

ollama serve

# Remember to pull the llama model in another terminal:
ollama pull llama3.2:3b

Running the program (Streamlit UI)

1. Navigate to the root directory /rag6
2. Use conda to activate the virtual environment and install dependencies

• conda activate rag6

3. Use the command streamlit run streamlit_app.py to run the UI
4. For each dataset, the model will take time to initialize and embed the vectors.
5. After the initialization, ask you question.