Team Collaboration

Multi-Database Team Studio

For design studios, gaming companies, and visual agencies with multiple artists working on shared network storage (NAS, SMB, NFS), Berry AI Studio can scale beyond local SQLite into a Multi-Database Team Collaboration Studio.

---

1. The Multi-Database Architecture

Berry provides an asynchronous StorageEngine abstraction layer supporting three backends:

Backend	Recommended Team Size	Concurrency Model	Performance Profile
SQLite (Default)	1 user per library	Single-writer / Multi-reader WAL	<0.5 ms latency on local NVMe SSDs.
MySQL 8.0+ / MariaDB	2 to 50+ concurrent users	Row-level locking & ngram full-text indexing	Sub-5ms queries on shared 500,000+ asset libraries.
PostgreSQL 14+	2 to 100+ concurrent users	MVCC, tsvector GIN indexes & LISTEN/NOTIFY	Low-latency real-time collaboration with broadcast events.

graph TD
    NAS[(Shared Studio NAS: SMB / NFS / WebDAV)]
    DB[(Central Studio DB: PostgreSQL 14+ / MySQL 8+)]

    subgraph Workstation A (Windows)
        A_UI[Berry UI]
        A_Thumb[Local NVMe Thumbnail Cache]
        A_UI --- A_Thumb
    end

    subgraph Workstation B (macOS)
        B_UI[Berry UI]
        B_Thumb[Local NVMe Thumbnail Cache]
        B_UI --- B_Thumb
    end

    A_UI -->|Z:\ai_vault| NAS
    B_UI -->|/Volumes/ai_vault| NAS

    A_UI <-->|OCC Version Checks & Sync| DB
    B_UI <-->|OCC Version Checks & Sync| DB

Loading

---

2. Cross-Platform Storage Root Mapping

A major challenge in cross-platform studios is that Windows, macOS, and Linux use different path formats for the same shared network folder:

• Windows: Z:\ai_vault\2026\character_01.png
• macOS: /Volumes/ai_vault/2026/character_01.png
• Linux: /mnt/nas/ai_vault/2026/character_01.png

How Berry Solves This:

1. Platform-Agnostic Root UUIDs: Berry generates a universal UUID for the shared storage root (stored in the database table storage_roots).
2. Normalized URIs: In the database, paths are stored platform-agnostically:

   berry://550e8400-e29b-41d4-a716-446655440000/2026/character_01.png
   

3. Client-Side Mount Mapping: In Settings > Team & Collaboration, each artist maps the Root UUID to their local operating system mount path. Berry dynamically translates paths on the fly, so an artwork tagged by an artist on macOS is instantly accessible to an artist on Windows.

---

3. Optimistic Concurrency Control (OCC)

When multiple team members rate, tag, or curate the same collection simultaneously, Berry prevents data corruption using Optimistic Concurrency Control (OCC):

• Every asset record includes a row-level version column.
• When an artist updates a rating, Berry submits set_file_rating_occ(file_id, new_rating, expected_version).
• Conflict Policies:
  • Ratings & Hero Covers: Last-Write-Wins (LWW) with immediate UI update.
  • Tags & Albums: Set-Union merge (if Artist A adds tag "Character" and Artist B adds tag "Concept", both tags are preserved).
  • Deletions: Soft-deletion status flags prevent phantom recreation.

---

4. Real-Time Collaboration Sync Tiers

Berry synchronizes state changes across the team using a 3-tier sync architecture:

1. Tier 1: Change Log Polling (Default / Zero-DevOps):
   • Berry queries the change_log journal table every 3 seconds for new transaction IDs. Requires no special server setup.
2. Tier 2: PostgreSQL LISTEN / NOTIFY (<50 ms Latency):
   • When using PostgreSQL, Berry establishes an asynchronous notification channel. Changes made by any workstation are broadcast to other clients in under 50 ms.
3. Tier 3: Distributed WebSocket Hub:
   • Optional lightweight broadcast service for larger enterprise studios.

---

5. Client-Side On-Demand Thumbnail Caching

Loading thumbnails across a 1Gbps or 10Gbps studio network can saturate shared bandwidth:

• Berry stores downscaled WebP thumbnails on each workstation's local NVMe SSD (thumbnails/).
• When an artist browses the shared library, thumbnails are generated and cached locally on demand.
• This keeps network traffic near zero during fast gallery scrolling, leaving shared NAS bandwidth free for high-resolution renders and model training.

---

6. Migration Wizard: SQLite to MySQL / PostgreSQL (MigrationWizardModal.vue)

If you started with a single-user SQLite library and want to upgrade to a shared team database:

1. Open Settings > Team & Collaboration.
2. Click "Launch Central Database Migration Wizard".
3. Berry inspects your local SQLite database, lets you choose MySQL 8.0+ or PostgreSQL 14+, and generates dialect-optimized DDL schemas and transactional SQL batch migration files.
4. Execute the migration script on your database server to transition your studio's library.