Restructure ROADMAP.md to reflect minimal-first implementation strategy

[Copilot]

Clarifies that DatabaseLoader is the single P0 blocker preventing production deployment. Restructures Phase 4 service implementation timeline to emphasize critical path vs. incremental upgrades.

Changes

Main ROADMAP.md

Added "Minimal Implementation Strategy" section

• Explains stub → MVP → production upgrade path for services
• Documents that dev-plugin provides in-memory fallbacks for all non-critical services
• Identifies DatabaseLoader as the only blocker for production metadata persistence

Restructured Phase 4 into time-boxed subphases

• Phase 4a (P0, Weeks 1-2): DatabaseLoader implementation in packages/metadata/src/loaders/database-loader.ts
  • Implements MetadataLoader interface with datasource: protocol
  • Maps to sys_metadata table CRUD operations via IDataDriver
  • Enables platform-level metadata editing, user overlay persistence, multi-instance sync
• Phase 4b (P1, Weeks 3-4): Infrastructure service upgrades
  • Redis, BullMQ, cron, S3 adapters (services already have memory fallbacks)
• Phase 4c (P1, Weeks 5-6): Communication & search services
  • i18n, realtime, search, notification (minimal viable implementations)
• Phase 4d (P2, Future): Business logic services
  • Automation, workflow, GraphQL, AI, analytics

Updated metadata ROADMAP.md

• Added "Last Updated: 2026-02-18" for consistency

Context

Auth service (✅ via better-auth) and UI service (⚠️ deprecated, merged into IMetadataService) statuses were already correctly reflected in Contract Implementation Matrix—no changes needed.

Original prompt

Context

Based on a thorough audit of the current codebase, the actual implementation status is:

Already Complete (DO NOT change these statuses)

• auth ✅ — Fully integrated with better-auth v1.4.18 (packages/plugins/plugin-auth/). Supports email/password, OAuth, 2FA, passkeys, magic links, ObjectQL-based DB persistence. 11/11 tests passing.
• ui ✅ — Merged into metadata service. IUIService is @deprecated (see packages/spec/src/contracts/ui-service.ts). MetadataManager provides getView(), listViews(), getDashboard(), listDashboards() directly. The API_IMPLEMENTATION_PLAN.md section 4.4 confirms: "P1: plugin-ui-api（已合并到 metadata API）".
• data ✅ — Full CRUD via ObjectQL (9 protocol methods)
• analytics ✅ — 2 protocol methods implemented
• metadata ⚠️ Framework — Works with MemoryLoader, FilesystemLoader, RemoteLoader. 30+ methods in IMetadataService implemented. Overlay system works in-memory.

The Critical Gap: DatabaseLoader

The single P0 blocker preventing frontend from moving to production mode is metadata DatabaseLoader (src/loaders/database-loader.ts in packages/metadata/).

Current metadata loaders:

• MemoryLoader (memory:) — ✅ Implemented
• FilesystemLoader (file:) — ✅ Implemented
• RemoteLoader (http:) — ✅ Implemented
• DatabaseLoader (datasource:) — 🔴 NOT IMPLEMENTED (this is the gap)

All spec schemas are already defined:

• MetadataRecordSchema — DB record envelope (metadata-persistence.zod.ts)
• MetadataManagerConfig.datasource — datasource name reference
• MetadataManagerConfig.tableName — defaults to sys_metadata
• MetadataManagerConfig.fallback — 'filesystem' | 'memory' | 'none'
• MetadataLoaderContract protocol datasource:
• IDataDriver interface — find, findOne, create, update, delete

Task: Update ROADMAP.md

Update the main ROADMAP.md file at the repository root to reflect the minimal-first implementation strategy. Key changes:

1. Update Phase 4 (Service Implementations — currently 🔴 In Progress)

Restructure to emphasize the minimal critical path:

Phase 4a: Metadata Persistence (P0 — Weeks 1-2)

• Implement DatabaseLoader in packages/metadata/src/loaders/database-loader.ts
  • Implement MetadataLoader interface with protocol datasource:
  • Accept IDataDriver instance via kernel DI
  • Map to sys_metadata table CRUD operations
  • Support scope filtering (system/platform/user)
  • Auto-create sys_metadata table on first use
• Integrate into MetadataManager (auto-configure when config.datasource is set)
• Persist overlay customizations to database
• This unblocks: platform-level metadata editing, user overlay persistence, multi-instance sync

Phase 4b: Infrastructure Service Upgrades (P1 — Weeks 3-4)

• Upgrade cache from in-memory fallback → Redis plugin
• Upgrade queue from in-memory fallback → BullMQ plugin
• Upgrade job from in-memory fallback → real cron scheduler
• Implement file-storage MVP (local FS driver + S3 driver)

Phase 4c: Communication & Search Services (P1 — Weeks 5-6)

• i18n — Map-backed translation service with locale resolution
• realtime — WebSocket/SSE push (replace Studio's setTimeout hack)
• search — Full-text search (in-memory first, then Meilisearch/Elastic driver)
• notification — Multi-channel notification engine

2. Correct auth status

Auth should be shown as ✅ Complete (better-auth integration) wherever it currently shows as 🟡 In Development or ❌ Plugin Required.

3. Correct ui status

UI service should be shown as ✅ Merged into Metadata wherever it currently appears as ❌ Plugin Required.

4. Update Contract Implementation Matrix

Ensure the matrix reflects:

• auth: ✅ (better-auth)
• ui: ✅ (merged into metadata, @deprecated as standalone)

5. Add "Minimal Implementation Strategy" section

Add a section near the top explaining the philosophy:

• Implement the smallest possible working version first
• DatabaseLoader is the single P0 because it's the only thing blocking production use
• All other services already have in-memory fallbacks via dev-plugin
• Each service can be independently upgraded from stub → MVP → production

Important Notes

• Keep all existing completed phases intact (Phase 1, 2, 3, 5 etc.)
• Do NOT remove or modify already-completed items
• Update the "Last Updated" date to 2026-02-18
• Make sure to run tests after changes to verify nothing breaks
• The file is ROADMAP.md at the repository root
• Also check and update packages/metadata/ROADMAP.md if it has stale information about auth/ui status

This pull request was created from Copilot chat.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
objectstack-play	Canceled		Feb 18, 2026 11:53am
spec	Ready	Preview, Comment	Feb 18, 2026 11:53am

[Copilot]

Pull request overview

Updates the project roadmap documentation to emphasize a minimal-first service delivery strategy and to clarify that metadata database persistence (DatabaseLoader) is the key production blocker.

Changes:

• Added a “Minimal Implementation Strategy” section to the root ROADMAP.md.
• Restructured Phase 4 into time-boxed subphases (4a–4d) centered on the DatabaseLoader critical path.
• Updated “Last Updated” dates in both roadmap documents for consistency.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
ROADMAP.md	Adds minimal-first strategy section and reorganizes Phase 4 into subphases with a clearer critical path.
packages/metadata/ROADMAP.md	Adds a “Last Updated: 2026-02-18” header line.

[Copilot]

Phase 4a tests mention “Integration tests with MemoryDriver”, but the in-repo driver class exported by @objectstack/driver-memory is InMemoryDriver. Consider renaming this roadmap item to InMemoryDriver (or clarifying what “MemoryDriver” refers to) to avoid confusion when implementing the tests.

Suggested change

	- [ ] Integration tests with `MemoryDriver`
	- [ ] Integration tests with `InMemoryDriver`

[Copilot]

Phase 4c is labeled “P1 — Weeks 5-6”, but the table includes INotificationService as P2. This makes the timeline ambiguous—either move notification into Phase 4d (P2) or update Phase 4c’s scope/priority so the phase only contains P1 work.

Suggested change

	| `INotificationService` | **P2** | `@objectstack/service-notification` | Email adapter (console logger in dev mode) |
	| `INotificationService` | **P1** | `@objectstack/service-notification` | Email adapter (console logger in dev mode) |

[Copilot]

Phase 4’s stated goal is to implement the remaining service contracts, but Phase 4b only enumerates cache/queue/job/storage upgrades. Since ISchemaDriver is still listed as spec-only in the Contract Implementation Matrix, consider explicitly placing it into a Phase 4 subphase (or calling out that it’s deferred to Phase 6) so the critical path for objectstack migrate/DDL work isn’t lost in the plan.