right now concore run --type docker gives us generated study artifacts but we still have to manually wire container startup. this is the missing piece.
We should add compose generation directly in run flow (opt-in flag is fine, ex --compose), write docker-compose.yml into output dir, one service per node, predictable names, mounts for study dirs, and just let users do docker compose up.
not asking to touch old backend internals. keep this additive in CLI layer.
What I’d expect done:
--compose only with --type docker
compose file generated in output dir
works for multi-node graph (mixed language too)
basic tests for single/multi node
small README snippet showing command + run
right now concore run --type docker gives us generated study artifacts but we still have to manually wire container startup. this is the missing piece.
We should add compose generation directly in run flow (opt-in flag is fine, ex --compose), write docker-compose.yml into output dir, one service per node, predictable names, mounts for study dirs, and just let users do docker compose up.
not asking to touch old backend internals. keep this additive in CLI layer.
What I’d expect done:
--compose only with --type docker
compose file generated in output dir
works for multi-node graph (mixed language too)
basic tests for single/multi node
small README snippet showing command + run