Inspiration
Like most people, I have years of files spread across folders: reports in one place, photos in a phone backup, voice notes in Downloads, videos on an old drive. When I want everything about one project or trip, search can find files by name, but not the story behind them. I wanted to ask one question and get the whole thing back.
What it does
Life Chapters is a Skill for UGREEN AI NAS. You ask something like "Show me everything from my 2025 college project" and give a few hints: a date range, a folder, the people involved, or a few keywords. Life Chapters then:
- finds the related documents, photos, recordings and videos on your NAS
- reads the documents and, if you allow it, transcribes the recordings
- writes a one-page chapter: an overview, a dated timeline, the people involved, key facts and a full file index
- after you approve, tags every file, creates a photo album and saves the chapter Because of the tag, you can reopen the whole chapter later with one sentence. Nothing leaves your NAS.
How we built it
A UGREEN Skill is not an app. It is a set of instructions that the NAS's own AI agents follow. I used the official Uliya Skill Kit to check that every step maps to a real NAS capability, then designed the workflow around four agents:
- file-agent: searching files and tagging them
- photo-agent: searching photos by people, place, date and text, and creating the album
- document-agent: reading documents and writing the chapter
- knowledge-agent: transcribing recordings and saving the result Each step passes only real results to the next one. Anything that changes data (tags, albums, saving) waits for the user's confirmation. The Skill passes the kit's official checker with zero errors and zero warnings.
Challenges we ran into
My first idea was a developer tool with code search and database backups. After reading the Skill Kit, I realised Skills can only use the capabilities the NAS provides, with no scripts or servers. So I started over and focused on what the NAS does well: files, photos, documents and voice. The other challenge was honesty. When nothing matches, the Skill has to say so and stop, not show "similar" files. I designed the empty and partial results carefully so the user always knows what was found, what was read and what was skipped.
Accomplishments that we're proud of
- A single, simple conversation designed to combine four different NAS agents
- Using a file tag as a lasting link, so a one-off search becomes something you can come back to
- A workflow that never moves, renames or deletes your files
What we learned
Designing for an AI agent is a lot like writing for a person: every instruction has to make sense on its own, with clear limits and a clear answer for every outcome, including failure.
What's next for Life Chapters
- Testing on a real UGREEN AI NAS in Round 2 and tuning the search for each type of hint
- Suggesting chapters automatically, for example: "You have 300 photos from December 2025, want a chapter for it?"
- Year-in-review chapters that collect a whole year
Built With
- ai-agents
- claude-code
- markdown
- ugos
- ugreen-ai-nas
- uliya-skill-kit
