chore(infra): replace COMPOSE_FILE chain with profiles and sentinels

[geovannewashington]

What changed?

Replaces the manual COMPOSE_FILE string concatenation in bin/utils
with a new bin/docker-compose wrapper. bin/utils is deleted.

Env loading is now delegated to Compose via COMPOSE_ENV_FILES.
Feature flags are translated to COMPOSE_PROFILES for add-only
services. Overlay files are loaded via include: sentinels in
docker-compose.yml. EXTRA_COMPOSE_FILE in the Makefile is renamed
to SHELLHUB_INCLUDE_EXTRA to follow the same convention.

Why

Part of #111. Steps 2, 3 and 4 from the migration plan are collapsed
into a single PR since they are too coupled to split safely.

How to test

1. ./bin/docker-compose up -d - confirm all services healthy
2. ./bin/docker-compose down, set SHELLHUB_DATABASE=mongo in
   .env.override, ./bin/docker-compose up -d - confirm mongo
   present, postgres absent
3. Set SHELLHUB_DATABASE=migrate - confirm both databases present
4. Set SHELLHUB_DATABASE=invalid - confirm ❌ error and exit
5. Set SHELLHUB_ENV=development - confirm agent present

[github-actions]

Claude finished @geovannewashington's task in 1m 56s —— View job

---

Code Review Complete

Reviewed 8 files across code quality, security, testing, language patterns, and architecture. 1 inline issue posted (high severity).

The overall approach is solid — replacing COMPOSE_FILE chaining with include: sentinels and profiles is cleaner and lets Docker Compose handle env loading via COMPOSE_ENV_FILES. The required: false on depends_on for database services correctly prevents startup failures when only one database profile is active. Cross-repo compatibility with cloud/ is confirmed intact.

Issue found

bin/setup was not updated as part of this PR but still sources the deleted bin/utils (line 3) and calls exit_if_not_running (line 29). It will crash on any invocation. See inline comment for details.

---

If you've addressed the feedback and want a new review, tag @shellhub-io/admin and a team member can trigger it.

[otavio]

/review

[github-actions]

Claude finished @otavio's task in 2m 25s —— View job

---

Code Review Complete

Reviewed 6 files across code quality, security, testing, language patterns, and architecture. 1 inline issue posted (critical severity, 2 locations).

The overall approach is clean — replacing COMPOSE_FILE chaining with include: sentinels and COMPOSE_PROFILES is a solid simplification. docker-compose.override.yml support is preserved via Compose's native auto-merge (no COMPOSE_FILE or -f set). Cross-repo compatibility with cloud/ is intact. The peek() function safely reads env values with hardcoded caller inputs. The previous review's bin/setup issue has been addressed.

Issue found

grep -c 'true' exits with code 1 when zero lines match (POSIX behavior). Under set -eu (newly added in this PR), the SERVICES=$(...) assignment propagates that exit and kills the script before the "ShellHub is not running" error message is reached. Affects bin/cli:8 and bin/setup:32. Fix: append || true to the pipeline inside the command substitution.

---

If you've addressed the feedback and want a new review, tag @shellhub-io/admin and a team member can trigger it.

[github-actions]

CRITICAL: grep -c exits 1 when count is 0, killing the script under set -e before the error message prints.

grep -c 'true' returns exit code 1 when zero lines match (POSIX behavior — exit status is based on whether any lines matched, regardless of -c). Since set -eu is active on line 3, the command substitution SERVICES=$(...) propagates that non-zero exit, and the shell terminates immediately at this line. The [ "$SERVICES" -le 0 ] guard and the "ShellHub is not running" error message on line 9 are never reached.

The old bin/utils had the same pipeline inside exit_if_not_running, but was safe because bin/utils used set -o allexport — not set -e.

Same issue in bin/setup:32.

Suggested change

	SERVICES=$(docker compose ps -q | grep -v INFO: | xargs docker inspect -f '{{.State.Running}}' 2>/dev/null | grep -c 'true')
	SERVICES=$(docker compose ps -q | grep -v INFO: | xargs docker inspect -f '{{.State.Running}}' 2>/dev/null | grep -c 'true' || true)

[github-actions]

CRITICAL: Same grep -c / set -e issue as bin/cli:8.

See the comment on bin/cli:8 for full details. The script silently exits instead of printing the "not running" error message.

Suggested change

	SERVICES=$(docker compose ps -q | grep -v INFO: | xargs docker inspect -f '{{.State.Running}}' 2>/dev/null | grep -c 'true')
	SERVICES=$(docker compose ps -q | grep -v INFO: | xargs docker inspect -f '{{.State.Running}}' 2>/dev/null | grep -c 'true' || true)

[otavio]

/review

[github-actions]

Claude finished @otavio's task in 4m 32s —— View job

---

Code Review Complete

Reviewed 6 files across code quality, security, testing, language patterns, and architecture — no issues found. The code looks good as-is.

The overall approach is solid — replacing COMPOSE_FILE chaining with include: sentinels and COMPOSE_PROFILES is a clean simplification. The previous review's issues (bin/setup sourcing deleted bin/utils, grep -c under set -e) have been correctly addressed. Cross-repo compatibility with cloud/ is intact. The docker-compose.override.yml auto-merge behavior is preserved since neither COMPOSE_FILE nor -f flags are set. The peek() function safely reads env values with hardcoded caller inputs, and the eval usage is not exploitable.

---

If you push additional changes and want a new review, tag @shellhub-io/admin and a team member can trigger it.