Family Biographer

Autonomous private family memory chronicler for UGREEN AI NAS that curates photos, extracts visual context locally, and writes narrative journals with zero cloud exposure.

Comment

Inspiration

Modern families capture thousands of intimate moments every year, storing gigabytes of vacation photos and milestone videos across local NAS folders. Unfortunately, these memories often sit forgotten in "cold storage"—buried in deeply nested directory trees. At the same time, uploading personal family albums to public cloud AI platforms poses significant privacy risks, recurring subscription fees, and massive bandwidth costs.

We were inspired to build Family Biographer to give UGREEN AI NAS an empathetic, observant memory. By leveraging on-premise compute and storage, we wanted to transform cold media backups into living, beautifully written family chronicles without a single byte of private imagery ever leaving the home.

What it does

Family Biographer is an autonomous, privacy-first family memory chronicler running natively on UGREEN AI NAS via UGOS.

  • Natural Prompt Discovery: Users trigger the skill with simple conversational requests (e.g., "Chronicle our Kyoto vacation from last spring").
  • Local Multimodal Synthesis: It queries stored media via date and geo-metadata, chronologically clusters events, and inspects key anchor scenes locally to detect landmarks, visible signage, and emotional context.
  • Narrative Chronicle Generation: It synthesizes rich, narrative journal chapters in formatted Markdown and curates matching smart albums in the native UGOS gallery.
  • Strict Human-in-the-Loop Safety: The agent never creates albums or writes files autonomously. It always halts at an explicit confirmation checkpoint, presenting a draft summary for the user to approve or refine first.
  • Anti-Hallucination Fallback: If photos lack GPS metadata or no media matches the query, the skill handles it gracefully without fabricating dates or locations.

How we built it

  • Uliya Skill Standard: Designed under the official UGREEN Uliya Skill Kit specification, orchestrating deterministic workflows through SKILL.md without disallowed binaries or hidden scripts.
  • Native UGOS Capabilities: Bound directly to supported system capabilities, including albums.search for metadata queries, media.describe for local scene understanding, albums.create for gallery organization, and files.write for persisting Markdown journals.
  • Deterministic Pipeline: Implemented sequential data handoffs: intent parsing $\rightarrow$ spatial/temporal clustering $\rightarrow$ anchor visual captioning $\rightarrow$ narrative synthesis $\rightarrow$ safety gate $\rightarrow$ persistence.
  • Modular References: Created dedicated reference guidelines (chronicle-template.md and prompt-personality.md) to maintain a warm, reflective storytelling voice.

Challenges we ran into

  • Graceful Degradation for Missing Metadata: Many mobile photos strip EXIF GPS coordinates during transfer. We had to design strict deterministic fallback logic that sorts strictly by timestamp and flags locations as "Unspecified" rather than letting the LLM hallucinate locations.
  • Strict Guardrail Orchestration: Ensuring the agent never executes destructive writes or unwanted album creation required designing airtight confirmation gates that cleanly exit if the user declines.
  • Edge-Case Safety: Building robust zero-result handlers so the model reports missing media clearly and halts immediately, rather than pulling unrelated photos.

Accomplishments that we're proud of

  • Zero-Cloud Exposure: Achieving end-to-end memory curation, visual description, and story generation entirely on local hardware, ensuring 100% family data sovereignty.
  • Fully Compliant Package: Delivering a clean, standardized Uliya Skill package free of placeholders, TODOs, or proprietary cloud API dependencies.
  • Bridging Storage & Emotion: Turning static raw storage volumes into an engaging, structured family heirloom.

What we learned

  • The power of on-device AI NAS: how local compute combined with terabytes of direct storage unlocks privacy-centric multimodal workflows that are impractical or cost-prohibitive on the cloud.
  • Best practices in authoring production-grade agentic skill packages with strict permission handling and deterministic step execution.

What's next for Family Biographer

  • Round 2 Hardware Testing: Benchmarking batch vision indexing and native album creation hooks directly on the remote UGREEN AI NAS iDX6011 Pro hardware.
  • Local Audio Integration: Adding on-device Whisper transcription to incorporate spoken voice memos and video dialogue into chronicle chapters.
  • Time Capsule Chat: Implementing interactive semantic search, enabling family members to converse with their chronicles (e.g., "When was the first time Sarah saw the ocean?").

Built With

Share this project:

Updates

Submission history