Nillion Movement POC

Introduction

This is the Nillion Movement POC

Prerequisites

1. Ollama: Install Ollama locally. Follow the Ollama Installation Guide.
2. Brew: Install Homebrew. Follow the Homebrew Installation Guide.
3. Docker: Install Docker. Follow the Docker Installation Guide.

Python and SQLiteVec configuration for compatibility on Mac

(Due to this issue)

1. Install SQLite for Mac

brew install sqlite

2. Install Python via Homebrew

(Follow the workaround outlined in the SQLiteVec documentation)

brew install python@3.12

3. Verify the Python Version

python3.12 --version

The Python version should display as 3.12.*.

Finetune models for the classifier

1. Change directory to the nillion_movement_poc/agent_modules directory

2. Make sure poetry uses the correct Python version

poetry env use python3.12

3. Run the poetry install command

poetry install

4. Finetune the blockchain action classifier

poetry run python agent_modules/classifier/blockchain_action/finetune_blockchain_actions.py

5. Finetune the domain classifier

poetry run python agent_modules/classifier/domain/train_domain_classifier_distilbert.py

Run the poetry install to create the virtual environment and install the dependencies

1. Change directory to the nillion_movement_poc/mighty_agent directory

2. Make sure poetry uses the correct Python version

poetry env use python3.12

3. Run the poetry install command

poetry install

OpenSearch data setup

1. Start the OpenSearch container (run from the nillion_movement_poc/mighty_agent directory):

docker compose -f opensearch/docker-compose.ci.yml up -d

2. Insert the data into OpenSearch by running the /mighty_agent/opensearch/opensearch_insert_docs.ipynb notebook.

Make sure you are choosing the correct interpreter in the notebook by following the steps below:

• Check the current python virtual environment:

poetry env info

Sample output:

Virtualenv
Python:         3.12.9
Implementation: CPython
Path:           /Users/nguyentruong/Library/Caches/pypoetry/virtualenvs/mighty-agent-QIDbtWhq-py3.12
Executable:     /Users/nguyentruong/Library/Caches/pypoetry/virtualenvs/mighty-agent-QIDbtWhq-py3.12/bin/python
Valid:          True

Base
Platform:   darwin
OS:         posix
Python:     3.12.9
Path:       /opt/homebrew/opt/python@3.12/Frameworks/Python.framework/Versions/3.12
Executable: /opt/homebrew/opt/python@3.12/Frameworks/Python.framework/Versions/3.12/bin/python3.12

3. Choose the interpreter that matches the virtual environment showed above (which is /Users/nguyentruong/Library/Caches/pypoetry/virtualenvs/mighty-agent-QIDbtWhq-py3.12 in my case)

4. Run the notebook with the correct interpreter (with VSCode-based notebook, click the Run All button):

Install ollama and required models

1. Make sure you have ollama installed

2. Install the required models

ollama run llama3.2:3b

Run the chat application

poetry run chainlit run app.py