[chillwur]

Problem

apps/backend/src/config/env.ts read process.env with hand-rolled required()/optional() helpers. Missing or malformed variables (e.g. a non-numeric PORT, or an unrecognized STELLAR_NETWORK) weren't caught in a structured way, so bad configuration could surface later as a confusing runtime crash instead of a clear failure at boot.

Solution

• Replaced the ad-hoc env reads with a zod schema (envSchema) covering every variable documented in apps/backend/.env.example: NODE_ENV, HOST, PORT, DATABASE_URL, REDIS_URL, the WHATSAPP_* vars, the STELLAR_* vars, HASHPORT_CONTRACT_ID, SIGNER_SECRET_KEY, and WALLET_ENCRYPTION_KEY.
• The schema distinguishes required-in-production from optional-in-dev vars via a superRefine check keyed on NODE_ENV: storage/WhatsApp/Stellar/secret vars stay optional for local development (preserving the existing behavior where the backend boots without WhatsApp/Stellar credentials), but become required once NODE_ENV=production.
• On validation failure, the backend now logs a readable, one-per-line list of exactly which keys are missing or invalid and exits non-zero (loadEnv() in env.ts), instead of failing later inside request handling.
• parseEnv() and formatEnvIssues() are exported as pure functions (no process side effects) so they're directly unit-testable.
• Added apps/backend/test/env.test.ts covering: default values with no vars set, a malformed PORT, an invalid STELLAR_NETWORK, an invalid NODE_ENV, missing production-required vars, a fully-populated production config, and the readable-error-list formatting.
• The public shape of env and secrets (including the lazy secret getters) is unchanged, so no call sites elsewhere in the backend needed to change.

closes #2