update ci to run tests - #35

• Trigger the GitHub Actions workflow on pull requests targeting master.
• Add a Python 3.11 setup step and install test dependencies.
• Run pytest in CI before the Docker build step.

graph TD
  A{{"Pull request to master"}} --> B["GitHub Actions workflow"] --> C["Setup Python 3.11"] --> D["Install deps"] --> E["Run pytest"] --> F["Docker build"]

  subgraph Legend
    direction LR
    _evt{{"Event"}} ~~~ _job["Workflow/Job"] ~~~ _step["Step"]
  end

Loading

The following are alternative approaches to this PR:

1. Split CI into separate workflows (tests vs build/push)

• ➕ Avoids running Docker build/push logic on PRs if not needed
• ➕ Clearer separation of responsibilities and permissions (e.g., no registry creds on PRs)
• ➕ Faster feedback by running only tests on PRs
• ➖ More workflow files to maintain
• ➖ Requires some duplication (checkout/setup steps) unless using reusable workflows

2. Use a matrix for Python versions (e.g., 3.10/3.11/3.12)

• ➕ Catches compatibility issues earlier
• ➕ Future-proofs releases across supported runtimes
• ➖ Longer CI runtime and higher GitHub Actions usage
• ➖ May require dependency pinning fixes

Recommendation: Current approach is a solid incremental improvement (adds PR gating with pytest) and is appropriate if Docker build should still occur on PRs. If the intent is purely test gating (and to avoid privileged build/push steps on PRs), prefer splitting into a dedicated test workflow and keeping build/push limited to pushes/tags.

Other (1) +16 / -0

.github-ci.ymlTrigger workflow on PRs and run pytest with Python 3.11 +16/-0

Trigger workflow on PRs and run pytest with Python 3.11

• Adds a pull_request trigger targeting master. Introduces steps to set up Python 3.11, install requirements plus pytest, and run pytest before the existing Docker build step.

.github/workflows/.github-ci.yml

Description

The workflow runs pytest after installing only requirements.txt and pytest, but the test suite
imports torch and pydantic, which are not listed in requirements.txt, so CI will fail during
test collection with ModuleNotFoundError. This makes the new PR test run non-functional as-is.

Code

.github/workflows/.github-ci.yml[R17-29]

+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+
+      - name: Install dependencies
+        run: |
+          pip install --upgrade pip
+          pip install -r requirements.txt
+          pip install pytest
+
+      - name: Run tests
+        run: pytest -q

Evidence

The workflow only installs requirements.txt plus pytest, but requirements.txt does not include
torch or pydantic, while the test files import them unconditionally. Therefore a clean CI runner
will fail on pytest collection due to missing modules.

.github/workflows/.github-ci.yml[17-29]
requirements.txt[1-22]
tests/test_batch_agg.py[1-4]
tests/test_pooling.py[1-4]
tests/test_longest_substring.py[1-4]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
CI installs `requirements.txt` + `pytest` and then runs `pytest -q`, but the repository tests import `torch` and `pydantic`. Since those packages aren’t installed by the workflow (and aren’t in `requirements.txt`), pytest will error during import/collection.

## Issue Context
Several tests under `tests/` have unconditional imports of `torch` and `pydantic`.

## Fix Focus Areas
- Add missing runtime test dependencies (at minimum `torch` and `pydantic`) to an installed requirements set.
- Prefer introducing a `requirements-dev.txt` / `requirements-test.txt` and installing it in CI (instead of mixing dev/test deps into prod deps if that’s undesirable).

### Suggested implementation
- Create `requirements-test.txt` (or `requirements-dev.txt`) containing at least: `pytest`, `torch`, `pydantic`.
- Update workflow to run:
 - `pip install -r requirements.txt -r requirements-test.txt`
 - then `pytest -q`

## Fix Focus Areas (code pointers)
- .github/workflows/.github-ci.yml[17-29]
- requirements.txt[1-22]
- tests/test_batch_agg.py[1-4]
- tests/test_pooling.py[1-4]
- tests/test_longest_substring.py[1-4]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

Adding a pull_request trigger causes the existing Docker Hub login/push steps to run on PRs, which
will fail on forked PRs (secrets unavailable) and can publish unreviewed PR code to
dchou1618/pyeng:latest for same-repo PRs. This creates both CI reliability failures and a
supply-chain/release integrity risk.

Code

.github/workflows/.github-ci.yml[R6-7]

+  pull_request:
+    branches: [ master ]

Evidence

The workflow is now configured to run on PRs, and the same job definition includes Docker Hub
authentication using secrets and a docker push to the :latest tag—so PR-triggered runs will
reach those deployment steps unless explicitly gated.

.github/workflows/.github-ci.yml[3-8]
.github/workflows/.github-ci.yml[31-38]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The workflow now triggers on `pull_request`, but the job still includes Docker Hub authentication and `docker push` to the `:latest` tag. This makes PR validation runs attempt deployment steps, which (a) fail on forked PRs due to missing secrets and (b) can publish unreviewed code to Docker Hub for same-repo PRs.

## Issue Context
The deployment steps already exist in the workflow; the PR adds the `pull_request` trigger, expanding when those steps execute.

## Fix Focus Areas
- Add an `if:` guard (or split jobs) so Docker login/push only runs on trusted events (e.g., `push` to `master`).
- Ensure PR runs only execute safe steps (checkout, install, test, build).
- Prefer `docker login --password-stdin` to avoid passing secrets via CLI args.

### Suggested implementation
- Add `if: github.event_name == 'push' && github.ref == 'refs/heads/master'` to **Authenticate** and **Push** steps, or move them into a separate job that only runs on push.
- Optionally tag images with `github.sha` and only update `latest` on push to `master`.

## Fix Focus Areas (code pointers)
- .github/workflows/.github-ci.yml[3-8]
- .github/workflows/.github-ci.yml[31-38]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools