Deployment

The simulation is embarrassingly parallel: each pod/container runs the same image against markets.csv to accumulate many independent market sessions. “Parallelism” at scale comes from running many identical instances, not from a single distributed job.

Docker

The image is built from deep_trader_tbse/Dockerfile:

$ docker pull armandcismaru/deeptrader:deeptrader2.5
$ docker run armandcismaru/deeptrader:deeptrader2.5

The container entry point is python3 tbse.py markets.csv.

Build-context caveat: the CMD runs tbse.py from the image working directory /app, and COPY . /app copies the build context’s root. Because tbse.py lives inside deep_trader_tbse/, the image must be built with that directory as the build context (docker build deep_trader_tbse/), not the repository root. See the repo-root AGENTS.md for the full list of Docker/dependency caveats (the Dockerfile installs dependencies inline rather than from requirements.txt, so versions can drift).

Kubernetes

market-simulations.yaml defines a batch/v1 Job:

$ kubectl apply -f market-simulations.yaml
$ kubectl get pods

It requests completions: 9 / parallelism: 9 — i.e. nine pods running the same image. Adjust completions/parallelism to scale; update the image: tag when you push a new version.

AWS provisioning (`configure_ec2.py`)

configure_ec2.py is a one-off bootstrap script that launches an EC2 instance (installing Docker + Kubernetes via user-data) and, optionally, creates a kops cluster. It is environment-specific — it contains hard-coded region, AMI, key-pair, and security-group identifiers tied to the original author’s AWS account, and much of it is commented out. Treat it as a historical reference, not a turnkey tool, and review it before running.

Credentials & buckets

The project uses the default AWS SDK credential provider chain (environment variables, shared profile, or instance role) — no credentials are hard-coded in the source. Copy .env.example to .env and set values as needed:

AWS_S3_INPUT_BUCKET — where training CSVs are read from for pickling (utils.pickle_s3_files).
AWS_S3_OUTPUT_BUCKET — where TBSE experiment outputs are uploaded (tbse.py).

Note: in the current source the S3 upload/download calls in tbse.py are commented out, so a default run writes results to local CSV files only.

Continuous integration

Three GitHub Actions workflows are defined in .github/workflows/:

Workflow	Trigger	What it does
`pylint.yml`	push, weekly cron, manual	Lints every `*.py` file with pylint.
`simulation-check.yml`	PRs to `main`, weekly cron, manual	Runs `tbse.py markets.csv` and asserts it prints `Done Now`.
`pages.yml`	push to `main` touching `docs/**`	Builds `docs/` into a static site with Jekyll and deploys it to GitHub Pages.

Dependabot (.github/dependabot.yml) opens weekly pip dependency-update PRs — which is the bulk of the recent commit history.

Enabling the documentation site (one-time)

pages.yml deploys via the GitHub Actions Pages flow, which requires the repo’s Pages source to be set accordingly: Settings → Pages → Build and deployment → Source = “GitHub Actions”. Until that is set once, the workflow builds but cannot publish. After enabling, every push to main that touches docs/** rebuilds and redeploys the site (served at https://<user>.github.io/DeepTraderX/).