TenantAtlas/specs/093-scope-001-workspace-id-isolation/contracts/cli.md
ahmido 92a36ab89e SCOPE-001: DB-level workspace isolation via workspace_id (#112)
Implements Spec 093 (SCOPE-001) workspace isolation at the data layer.

What changed
- Adds `workspace_id` to 12 tenant-owned tables and enforces correct binding.
- Model write-path enforcement derives workspace from tenant + rejects mismatches.
- Prevents `tenant_id` changes (immutability) on tenant-owned records.
- Adds queued backfill command + job (`tenantpilot:backfill-workspace-ids`) with OperationRun + AuditLog observability.
- Enforces DB constraints (NOT NULL + FK `workspace_id` → `workspaces.id` + composite FK `(tenant_id, workspace_id)` → `tenants(id, workspace_id)`), plus audit_logs invariant.

UI / operator visibility
- Monitor backfill runs in **Monitoring → Operations** (OperationRun).

Tests
- `vendor/bin/sail artisan test --compact tests/Feature/WorkspaceIsolation`

Notes
- Backfill is queued: ensure a queue worker is running (`vendor/bin/sail artisan queue:work`).

Spec package
- `specs/093-scope-001-workspace-id-isolation/` (plan, tasks, contracts, quickstart, research)

Co-authored-by: Ahmed Darrazi <ahmed.darrazi@live.de>
Reviewed-on: #112
2026-02-14 22:34:02 +00:00

49 lines
1.6 KiB
Markdown

# CLI Contract — 093 Workspace ID Backfill
This feature adds an operator-only Artisan command to backfill missing `workspace_id` on tenant-owned tables.
## Command
- Name (proposed): `tenantpilot:backfill-workspace-ids`
## Flags / Options (proposed)
- `--dry-run` (default: false)
- Prints counts per table and exits without writing.
- `--table=<name>` (optional)
- Restrict execution to a single table.
- `--batch-size=<n>` (default: 5_000)
- Batch size for updates (where chunking is used).
- `--resume-from=<cursor>` (optional)
- Resume from a saved cursor/checkpoint (implementation-defined).
- `--max-rows=<n>` (optional)
- Safety valve for partial runs.
## Safety + Observability
Execution strategy (queued):
- The command is a start surface only: authorize → acquire lock → create/reuse `OperationRun` → dispatch queued jobs → print a “View run” pointer.
- The backfill mutations MUST execute inside queued jobs (batch/table scoped) to support large datasets.
Safety + observability requirements:
- Must acquire a lock (cache/DB-backed lock) to prevent concurrent runs.
- Must create/reuse an `OperationRun` for visibility and progress tracking.
- Must write an `AuditLog` entry for start and end (outcome, counts, duration).
- Must abort and report when a tenant→workspace mapping cannot be resolved.
## Output
- Printed “Run started” summary:
- `OperationRun` identifier (or URL/route reference when available)
- jobs dispatched count
- selected tables / scope
- Per-table totals:
- scanned rows
- rows missing `workspace_id`
- rows updated
- Final summary + recommended validation SQL.