Running TradingAgents locally

Tauric TradingAgents is an open source project that uses several AI agents to work through investment research. The agents take on different roles: analysts, researchers, a trader, risk reviewers, and a portfolio manager. The project’s README explains how they work together. This post covers how I connected the framework to cloud and local models, with a rate limiter to keep requests under control.

The usual setup needs API keys for cloud models. I already had AI subscriptions and wanted to use whatever allowance I had left in those instead of paying separately for API calls. So I put a local model gateway on my MacBook. It lets me inspect the traffic and limit requests, and I can switch providers without changing the TradingAgents code.

What’s Hugo?

Hugo introduces itself as a “Fast & modern static site engine”, with the aim to make the web fun again. After ages of CGI scripts evolution, easy and attractive CMSes, it looks like static page are back in fashion. Initially made popular by Jekyll a couple of years before, Hugo is one of the last additions to the family. It’s open source, written in Go, with a handful of themes, it’s easy to install and use.

OpenShift is Red Hat’s Platform-as-a-Service (PaaS) that allows developers to quickly develop, host, and scale applications in a cloud environment. It has decent documentation and relatively steep learning curve. It also provides free hosting + free MySQL DB up to 1 GB, which is aweosome!

After installing Ruby and RHC - Openshift’s client tool, by following these steps, you can start creating a Java web application following these steps.

Here https://github.com/atsamour/Heroku-java-example you can find a sample Java web project for Heroku PaaS cloud. You can also run it locally with only JRE and Maven installed, as it doesn’t requre a seperate application container. It evaluates Heroku’s open source with embedded Tomcat.

To run locally

First you have to git pull the repo to your PC, then build the application running this command from the local project folder:

:::term
$ mvn package

And then you just run the application by using the java command: