[er-s-an]

Description

RAGAnythingConfig reads scalar environment defaults while its module is imported. Loading a .env file afterwards, as application code commonly does, therefore leaves parser, storage, concurrency and processing settings at stale defaults. For example, loading PARSER=docling, MAX_CONCURRENT_FILES=4 and ENABLE_IMAGE_PROCESSING=false after import still constructs a MinerU configuration with concurrency 1 and images enabled.

Use default factories for the scalar environment settings, matching the existing list settings. Each new configuration captures the environment at construction; explicit arguments still take precedence. Existing instances retain their values. Keep the legacy parse-method compatibility path unchanged.

Related Issues

Found during local investigation. Searches for configuration/environment defaults, default_factory, and dotenv found no equivalent pending fix or claimed issue. #315 concerns packaging and other reliability changes, not these defaults.

Validation

At upstream afb67c43d391b1b1008a6fee9f020a232067b9e5, Linux/Python 3.12:

• Both new regressions fail before the fix.
• 46 focused/core tests pass after the fix; full suite: 900 passed.
• Ruff formatting/checks and git diff --check pass.
• Separate final diff review checked constructor precedence, existing-instance stability, environment cleanup in tests, and unchanged legacy behavior.

Tests import the real config class and load an actual temporary dotenv file. They cover string/integer/boolean values and two configurations constructed under different environments. No live model, external parser executable, or end-to-end ingestion was exercised; other Python/OS combinations await CI.

Checklist

• Changes tested locally
• Code reviewed
• Class documentation updated
• Unit tests added

Additional Notes

AI assistance: OpenAI Codex investigated, implemented, tested and reviewed this change. No independent human review is claimed.