Doppler, direnv & injected environments

The default resolver writes the per-branch connection into your .env file, which covers most projects. But plenty of teams inject environment instead — Doppler, direnv, CI secrets. branchly is friendly to all of them, on both the input and the output side.

The input side: where branchly finds your connection

branchly derives everything from your existing DATABASE_URL — it keeps the host and credentials and swaps the database name. The catch is that your git hooks run in their own environment, which may differ from your shell. So during init, branchly detects how that value reaches your project and wires the hooks to match:

• .env files are auto-loaded, non-overriding — values already present in the process environment always win. So Doppler-, direnv-, or CI-injected values take precedence naturally.
• Doppler — detected from a doppler.yaml or a logged-in doppler CLI with a configured project. When chosen, init wraps the Git hooks in doppler run, so the hook sees the same secrets your app does.
• direnv — detected from an .envrc plus the direnv binary; init wraps the hooks in direnv exec.
• shell — a value already exported in your environment. Convenient, but a fresh shell, CI runner, or git hook may not have it, so branchly says so rather than assuming.

In an interactive terminal, init confirms the detected source with you (and lets you pick another); in CI or with --yes it takes the top match, falling back to .env. Either way it then verifies that DATABASE_URL actually resolves the way the hooks will see it — if it can’t, init tells you exactly what’s missing instead of writing a config that silently won’t work. You can pin the choice with branchly init --env=doppler (or direnv, env-file, shell), and re-check anytime with branchly doctor.

The output side: where your app finds the branch database

Option 1: file-based resolvers

The env-file resolver upserts the key into .env; the direnv resolver does the same with export lines in .envrc:

resolver: { use: 'direnv', file: '.envrc', key: 'DATABASE_URL' }

Option 2: branchly run for injected setups

When your environment is injected at launch time, a file write isn’t enough — the injector would overwrite it. branchly run solves the ordering: it provisions the current branch, then launches your command with the per-branch connection set last, so it wins:

doppler run -- branchly run -- npm run dev

Doppler injects your secrets, then branchly overrides just the database URL with the current branch’s. Works the same with any injector:

branchly run -- npm run dev          # plain
direnv exec . branchly run -- npm test

branchly run passes the child’s exit code through, so it composes fine with scripts and CI steps.