Env Loading Order and 403 Error | Generated by AI

Home 2026.07 English 中文

---

Here’s the env loading order in ww/env/__init__.py:

1. ~/.config/ww/.env — base defaults (override=False)
2. $CWD/.env — current working directory overrides (override=False) → you ran from /Users/lzwjava/projects/jekyll-ai-blog, so it checks jekyll-ai-blog/.env
3. /Users/lzwjava/projects/ww/.env — project root (override=False)
4. BASE_PATH env var — if set, loads $BASE_PATH/.env (override=True)

All three .env files load with override=False — meaning if a variable is already set from an earlier source, later files won’t override it. Only BASE_PATH uses override=True.

So if your ~/.config/ww/.env already has VISION_MODEL=google/gemini-2.5-flash-image, changing it in /Users/lzwjava/projects/ww/.env won’t take effect — the earlier load wins.

The 403 error itself is an OpenRouter ToS rejection for google/gemini-2.5-flash-image — Google models on OpenRouter require explicit opt-in to their terms of service on the provider’s side. This isn’t a .env issue per se, but switching to a non-Google vision model would fix it.

To fix either:

• Check your ~/.config/ww/.env — if VISION_MODEL is set there, change it there (it loads first, blocks the project root)
• Or set a different vision model like openai/gpt-4o-mini or anthropic/claude-3-5-sonnet-20241022 in whichever .env file loads first

Want me to check what’s in ~/.config/ww/.env?

---

Generated by AI. Curating and sharing still takes effort. If you find it useful, feel free to donate. WeChat: @lzwjava · X: @lzwjava · Say hi 👋

Back Donate