v0.3.1 - Bug fix

Bug Fix

Plugin: server not configured after selecting a recent host

When selecting a server from the recent hosts combo box in the Plugin Manager, the display string (in the form host:port - serverName) was being saved as-is to the preferences. This caused the host validation to fail in the Add Photos, Remove Photos, and Manage Tasks dialogs with a "Server not configured" error.

The host is now normalized before saving, and defensively extracted on read in all three dialogs to transparently recover from existing broken preferences.

Affected files: PluginInfoProvider.lua, AddPhotosDialog.lua, RemovePhotosDialog.lua, TaskDialog.lua

---

Full changelog: v0.3.0...v0.3.1

v0.3.0 - Multi-language support, stability, and plugin UX

We're releasing Photometoria v0.3.0, a broad stabilization and improvement release spanning both the API and the Lightroom plugin.

This milestone delivers multi-language tag generation, a more robust provider and catalog model, significant plugin UX improvements, and a set of targeted bug fixes and storage resilience enhancements.

What's included:

Multi-language tag generation (API + Plugin):

• Language parameter for tag generation — the API now accepts a target language per job (#66)
• Language selection in the New Job dialog (#67)
• Tag output switched from plain text to structured JSON for reliable parsing (#87)
• Validated against llava and qwen3-vl models across multiple languages (#65, #64, #88)

Provider and catalog model (API):

• Explicit provider required in job creation request — no more implicit fallback (#104)
• Provider name stored on the job for traceability (#103)
• New GET /api/catalogs endpoint to list available catalogs (#83)
• Tasks now associated with catalog identity to prevent cross-library mixing (#57)
• Configurable server name exposed on the info endpoint (#69)
• num_ctx (context window size) now configurable per Ollama model (#105)

Plugin UX improvements:

• Job creation on a subset of task photos — use the current Lightroom selection (#114)
• "Remove Photos" command added to the plugin (#113)
• Job timestamps and estimated time remaining shown in the plugin UI (#71)
• Option to remove existing keywords before applying AI-generated tags (#70)
• Recent hosts list in server connection preferences (#60)
• Last used model pre-selected when opening the New Job dialog (#79)
• Last active task updated when switching from the task popup (#81)

Bug fixes:

• Orphaned jobs no longer remain in memory after task deletion (#68)
• Photo store in-memory state correctly cleared on task deletion (#98)
• Job creation now returns HTTP 201 instead of 200 (#61)

Storage resilience (API):

• Filesystem consistency checks on startup (#17)
• Improved filesystem storage resilience (#24)
• Input validation added across API endpoints (#13)
• Effective configuration logged in human-readable format at startup (#74)

API cleanup:

• Deprecated GET /version and GET /api/models endpoints removed (#32)
• Re-uploading a photo replaces the existing entry instead of creating a duplicate (#58)

Next up (v0.4.0): TBD — see the issue tracker for planned work.

Feedback welcome.

v0.2.0 - Lightroom Plugin

We're releasing Photometoria v0.2.0, which delivers a full-featured Lightroom Classic plugin.

This milestone brings Photometoria from a standalone API into the photographer's actual workflow: photos can now be selected, uploaded, analyzed, and tagged entirely from within Lightroom Classic.

What's included:

Plugin (Lua):

• Task management — create, delete, rename, edit context
• Photo workflow — select, validate, export and upload photos to tasks
• Job lifecycle — start AI analysis, monitor real-time progress, cancel, retry, delete
• Keyword application — apply AI-generated tags as Lightroom keywords with confirmation dialog
• Show Photos — navigate to task photos in the Library module
• Progress indicators and job status icons in the UI

API enhancements (Rust):

• GET /api/info — server capabilities and status
• GET /api/providers and GET /api/models — provider and model discovery
• name field on tasks for user-friendly identification
• client_id on photos for client-side reconciliation
• Task endpoints now expose their associated jobs
• Photo summaries in task photo listing

Next up (v0.3.0): TBD — see the issue tracker for planned work.

Feedback welcome.

v0.1.0 - API Functional Completion

We're releasing the first functional version of the Photometoria REST API.

This milestone marks the completion of the core API layer: photos can now be uploaded, analyzed by a local AI model via Ollama, and the generated tags retrieved — all through a clean REST interface.

What's included:

• Full job lifecycle: create, monitor, cancel, retry, delete
• Worker pool with GPU-aware model switching
• Ollama integration with per-model prompt templates
• GET /api/models to list available vision models
• Delete guards to prevent accidental data loss during active jobs

Next up (v0.2.0): Work on the Lightroom Classic plugin begins.

This is an early release — the API is functional but not yet stable. Feedback welcome.