Pet Memory Atlas

Round 1 proposal | 15 November 2026 | NAS validation pending

1. Function

Turn authorized historical pet photo records into an evidence-linked memory archive with owner-confirmed names, shared-photo attribution and honest place summaries.

2. Target user and problem

Households with multiple pets forget which photos belong to whom and where memories happened. Category labels, incomplete searches and missing GPS can create misleading histories.

3. Proposed solution

One parent Skill coordinates five Markdown reference modules: discovery, identities, places, reporting and capability boundaries. It works without cameras and retains uncertainty rather than inventing recognition or travel details.

4. Core user flow

Confirm photo scope; recover a prior ledger if supplied; discover albums and search remembered descriptions; confirm names and shared-photo membership; collect available metadata; produce a retrospective; optionally save and reopen the archive.

5. Substantive difference

A single photo can belong to both confirmed pets without duplicate source counting. Owner corrections preserve stable identifiers. The archive separates matched, returned, read and visually inspected evidence, and distinguishes a coordinate from a confirmed venue.

6. NAS capabilities and workflow steps

Discovery and metadata use photo-agent with ug-photo albums, object-albums, search-classify, search-picture, search-screen and meta. Persistence uses knowledge-agent/ug-knowledge save-text. Recovery uses file-agent/ug-filesystem stat and cat. Optional vision-vlm obeys its original-user image-input contract.

7. Technical approach

Natural-language expert delegation uses official Uliya Kit 0.7.7-en and catalog 2.2.0. Inputs are authorized photo scope, owner confirmations and a real save destination. Photo access is read-only; saving requires authorization. The platform selects its configured AI; no external model is required. Verify actual model data flows before promising local-only processing. No shell, HTTP, private database or executable code is invoked.

8. Development status

The English orchestration Skill and five references are authored; the unchanged official checker reports zero errors and warnings. NAS execution remains untested. A separate local gallery prototype is excluded: its HTML themes, server and experimental recognition are not submitted Skill capabilities.

9. Final validation plan

On assigned NAS, core test one: discover an authorized two-pet album, confirm shared-photo attribution and verify source counts. Core test two: save a location-evidence archive and fully reopen it, preserving names and provenance. Edge test: missing GPS plus truncated search must retain unknowns, disclose incomplete coverage and invent no venue.

10. Limits and dependencies

Depends on available authorized photo, file and knowledge experts. No automatic individual-pet recognition, native identity renaming, website deployment, geocoding or monitoring is claimed. GPS and capture timezone may be absent. Real permissions, returned schemas, interruption recovery and save verification require NAS validation. No third-party code, weights or datasets are distributed or required by this ZIP. Sources, licenses and AI assistance are disclosed below and in the ZIP.

Sources, licenses and authoring assistance

UGREEN Starter Kit 0.7.7-en: official documentation/checker; UGREEN terms, not redistributed. Immich (AGPL-3.0) and Frigate (MIT core): background documentation only; no code or runtime integration included. Codex assisted Skill workflow drafting and review, rules research, disclosure editing, checks and packaging. PDF export uses ReportLab 4.4.9 (BSD) and pypdf 6.21.4 (BSD-3-Clause), excluded from the ZIP. Source and license links are in the ZIP's directly linked capability-boundary reference.

Built With

Share this project:

Updates

Submission history