bug: Missing automated test suite causes undetected regressions across critical modules #134

Problem

The reactome_chatbot repository currently has no automated test suite and no CI testing infrastructure. This creates several risks as the project scales:

• Contributors cannot easily verify their changes don't break existing behaviour
• Regressions in critical paths (config loading, retrieval logic) go undetected until runtime
• The absence of a test gate makes it harder to confidently merge PRs

Proposed Solution

Introduce a lightweight testing foundation using pytest, targeting the core modules most critical to chatbot operation:

1. util.config_yml — Validate that config.yml (and default fallbacks) parse correctly into Pydantic models, and that feature flags / usage limits evaluate as expected
2. retrievers.csv_chroma — Unit test utility helpers such as ChromaDB subdirectory discovery, ensuring retrieval scaffolding behaves correctly under edge cases

Implementation Plan

• Add a tests/ directory at the project root
• Add pytest and pytest-mock to pyproject.toml under [tool.poetry.group.dev.dependencies]
• Add a GitHub Actions workflow at .github/workflows/tests.yml that triggers on push and pull requests
• Write initial test coverage for the two modules above as a foundation others can build on

Example Test Structure

tests/
├── __init__.py
├── test_config_yml.py       # Config loading, defaults, validation
└── test_csv_chroma.py       # ChromaDB subdir discovery, edge cases

Impact

Related

This complements ongoing work in:

• feat: make LLM and embedding model configurable via YAML config #112 (configurable LLM/embedding via YAML — config tests will validate this)
• fix: replace mutable default arguments in Config methods #131 (mutable default args in Config — tests will catch this class of bug going forward)

Opening a PR with the initial tests/ scaffolding and GitHub Actions workflow.