Getting Started

NOTE: Currently, the project supports Python 3.10.9+. It is recommended to use a virtual environment or install a python supported version Python 3.10.9 Release

Install Dependencies

pip install -r requirements.txt

API key

By default variables.env will try to retrieve an OPENAI_API_KEY set on the local device. If you named your key differently or do not have it set up as an environment variable, you can manually set the API key in variables.env.

Primary Run Method: Run with Docker

Running the system with docker-compose will create a couple of clients and a Continuous Double Auction exchange. This option is recommended to conduct experiments as it uses all components of the system.

Install Docker Desktop

https://www.docker.com/get-started/ After installing Docker Dekstop, make sure it is running while performing any docker commands

Build images

The docker-compose file requires the images project and app to be built.
build project image

cd
docker build -t project .

build app image

cd user-interface
docker build -t app .

Note: You can skip these steps by running build.sh; however, it currently only works on linux

Running docker-compose

Make sure you're in the home directory where the docker-compose.yaml file is located. To start the system enter:

docker-compose up

This will create two clients:
client_0 is at http://localhost:5070
client_1 is at http://localhost:5071
To add more clients, update the docker-compose.yml by following the pattern for client_0 and client_1

To stop the system enter:

docker-compose down

Development Information

The following run methods are recommended approaches to debug/test the system in isolated modules.

Development Run Method # 1: Run locally without Docker(Recommended to test a single client)

To run the system without docker, you must manually start the Continuous Double Auction and clients. This option is recommended to test/debug flask_client.py without having to deal with the react frontend. Additionally, it is recommended to use this for a single client due to the structure of the logging system.

Run the CDA Exchange

python ./run_market_server.py

Run a CDA Client

Since the CDA Exchange is running locally on localhost:8090, you must specify that the host(CDA Exchange) is localhost and change the port of the client to one other than 8090. An example would look like of correctly running a client locally looks like:

python ./run_market_client.py --port 5001 --host localhost

If you want to run multiple CDA clients, run the same command in a different terminal and change the --local argument to a different unused port. To see the clients, visit http://localhost:<port>. In this example you can reach the client's Flask Endpoints by visiting http://localhost:5001.

run React frontend(optional)

If you want to use the React frontend with this run option run:

npm run dev

Note: You must manually check if the React frontend fetches from the correct client's Flask port

Development Run Method # 2: Run locally with clients in development mode

This option is a recommended way to test/debug the client.py. This does not run client's Flask endpoint or the React frontend; it only establishes a simple client-exchange interaction. run_market_client.py has an argument '--mode' or '-m' that's used to specify the mode to run a client. Using the term 'dev' will run a client in development mode. For more information on client configuration arguments visit: run_market_client.py. Similarly, visit run_market_server.py for Exchange configuration information.

Run the CDA Exchange

python ./run_market_server.py

Create a Client to Send Orders from the Terminal (Each Client Requires Own Terminal)

Creating a client this way will allow you to manually place orders without ChatGPT and Flask. As mentioned earlier, the CDA Exchange is running locally on localhost:8090 so we must specify the --host as localhost

python ./run_market_client.py -m dev --host localhost

Files

Folder: /Llama_index

• Contains llama_rag.py which currently asks for prompts and generates code based on them.
• Contains /system_data where we store descriptions of functions.
• Contains /helper_functions where the executable helper functions are stored.

Folder: /market_client

• Contains client.py which is the client class used in flask_client.py.
• Contains flask_client.py that connects a client to the market and with generated strategies.

Folder: /exchange_logging

• Contains exchange_loggers.py which contains the logging classes used by the client and the exchange server in client.py and exchange.py.

Project Backlog

Acknowledgements

This project is based on the previous work from Leeps-Lab/exchange_server.