run-backend

Rellm's backend is a Rust (cargo) server, driven entirely through backend/Makefile.

Starting the backend

cd backend && make local_minio_start run

• local_minio_start starts the rellm-dev-minio Docker container (creates it via local_minio_create if it doesn't exist yet).
• run is just cargo run.
• The server listens on ports 80 and 8000 locally.
• The user often has this running themselves already. It is safe to kill their dev server and restart it — the user has said this explicitly. To stop it: cd backend && make local_instances_stop (runs killall rellm), or find/kill whatever process is bound to ports 80/8000.

Rebuilding a frontend and serving it through the Rust server

These are the right targets when you want to validate a Tamagui or Elm change as served by the backend (as opposed to that frontend's own standalone dev server):

cd backend && make rebuild_tamagui_and_run   # rebuilds Tamagui FE, then cargo run
cd backend && make rebuild_elm_and_run       # rebuilds Elm FE, then cargo run

Prefer rebuild_tamagui_and_run over Tamagui's own dev server when sanity-checking a change — Tamagui's own dev loop (see the run-tamagui skill) is much slower to build/iterate with.

Other useful backend targets

• make build / make clean — plain cargo build/clean.
• make rebuild_protos — cargo clean -p prost-build then build, to force proto regeneration on the Rust side (see the rebuild-protos skill for the full multi-frontend proto rebuild).
• make local_db_create / local_db_drop / local_db_reset / local_db_connect — manage the local rellm_dev Postgres database.
• make local_minio_delete — stop and remove the MinIO container entirely.
• make test_authentication_local — resets the local DB, boots the server, and runs a grpcurl-based auth smoke test (create account / login, including expected-failure cases) against localhost:27707, then stops the server again.