Initial scaffold: Flask/SQLite backend, React/Vite PWA frontend

Backend: receipt upload -> LLM vision extraction (OpenAI-compatible,
provider-agnostic), item review/edit, per-group splitting against
Cospend projects/members, highlight+upload via WebDAV, public share
link, bill creation via Cospend's OCS API (verified against real
source, not just doc summaries).

Frontend: capture -> review -> group -> summary flow as an installable
PWA.

install.sh / run.sh (venv + npm, tmux session) instead of Docker, per
the ~/Projects/gain pattern.

backend/.env.example

@@ -0,0 +1,31 @@
+# Copy to .env and fill in. Never commit .env.
+
+# --- Nextcloud ---
+NC_BASE_URL=https://cloud.dominik-roth.eu
+NC_USERNAME=dodox
+# Settings -> Security -> "Devices & Sessions" -> create app password
+NC_APP_PASSWORD=
+# Folder (relative to the user's files root) receipt images get uploaded to.
+# Created automatically on first upload if missing (each path segment).
+NC_UPLOAD_FOLDER=Documents/Cospend/Assets
+
+# --- Cospend ---
+# Optional: pre-selects a project in the UI. The app lists all your Cospend
+# projects via the API (GET /ocs/v2.php/apps/cospend/api/v1/projects), so
+# this is just a convenience default, not required.
+COSPEND_PROJECT_ID=
+
+# --- Vision LLM (OpenAI-compatible chat completions API) ---
+# OpenAI example:
+#   LLM_BASE_URL=https://api.openai.com/v1
+#   LLM_MODEL=gpt-4o-mini
+# Gemini (OpenAI-compat layer) example:
+#   LLM_BASE_URL=https://generativelanguage.googleapis.com/v1beta/openai/
+#   LLM_MODEL=gemini-2.0-flash
+LLM_BASE_URL=
+LLM_API_KEY=
+LLM_MODEL=
+
+# --- Flask ---
+FLASK_SECRET_KEY=dev-change-me
+DATABASE_PATH=instance/wgbill.sqlite3