# Roadmap

| Phase                       | Focus                          | Why now?                                                                                    | Key deliverables                                                                                                              |
| --------------------------- | ------------------------------ | ------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- |
| **1. 2025-Q2 – “Bedrock”**   | Migrations + CI baseline       | We’ve landed versioned migrations and removed runtime column hacks – ensure it stays solid. | • CI job runs `cargo test` + `cargo sqlx migrate run --dry-run`                                                               |
| **2. 2025-Q2**              | Leaner FTS maintenance         | Per-row triggers don’t scale past ~100 k files.                                             | • Replace triggers with “dirty” flag + periodic rebuild <br>• Benchmark on 100 k files                                          |
| **2.1 2025-Q2** (adjunct)   | Dirty-row FTS + CI             | Prep for both scale and live-watcher—avoid full rebuilds on every change.                   | • `scan --dirty` reindexes only files whose SHA-256 hash has changed                                                          |
| **2.2 2025-Q2**             | Live file watching             | Offer a true “working-dir” mode—auto-scan on FS events, zero manual rescans.                | • New `marlin watch [dir]` subcommand using `notify` crate<br>• Incremental scans on create/delete/modify/rename events      |
| **2.3 2025-Q2**             | Self-pruning backups           | Instant protection and bounded storage—no manual snapshot cleanup.                          | • `marlin backup --prune <N>` flag<br>• Hook post-scan (when rows change) to auto-prune to latest 10<br>• Daily cron/automation to keep last 7 |
| **3. 2025-Q3**              | Content hashing & dedup        | Detect duplicates, enable future integrity checks.                                         | • SHA-256 in `files.hash` <br>• `scan --rehash` option                                                                          |
| **4. 2025-Q3**              | Tag aliases / canonicals       | Control tag sprawl before users accumulate thousands.                                       | • `canonical_id` enforcement <br>• `tag alias add|ls|rm` CLI                                                                   |
| **5. 2025-Q4**              | Search DSL v2                  | Power users want grouping, boolean ops, quoted phrases.                                     | • Replace ad-hoc parser with `nom` grammar <br>• Unit-tested examples                                                          |
| **6. 2025-Q4**              | Attribute templates            | Structured metadata unlocks real workflows.                                                 | • `templates` + `template_fields` tables <br>• Validation on `attr set`                                                        |
| **7. 2026-Q1**              | Dolphin read-only plugin       | Browse tags/attrs inside the default file manager.                                          | • Qt sidebar showing metadata                                                                                                 |
| **8. 2026-Q1**              | Full edit UI                   | After read-only proves stable, add editing.                                                 | • Tag editor widget, attribute dialog                                                                                         |
| **9. 2026-Q2**              | Multi-device sync              | Final frontier: optional sync/replication layer.                                            | • Choose between rqlite / Litestream / bespoke <br>• Read-only mode for network mounts                                         |

---

### Current sprint (ends **2025-06-01**)

1. **FTS rebuild prototype** – dirtied-rows approach, measure on 50 k files.  
2. **`backup --prune`** to keep only N most recent snapshots.  
3. Integration tests for tag/attr workflows on Windows via GitHub Actions.

---

### Development principles

* **Local-first** – every feature must work offline.  
* **Zero manual migrations** – shipping code *is* the migration.  
* **Instrumentation first** – every new command logs trace spans and timings.  
```