ArchiveWitness

Inspiration

Family photographs, handwritten letters, and scanned documents preserve valuable memories, yet their stories often become fragmented over time. A photograph may have no date. A letter may mention a place without explaining its significance. When families try to reconstruct these memories using conventional AI tools, plausible but unsupported details can enter the narrative. ArchiveWitness was conceived to address this problem: preserving family stories while keeping every factual claim connected to its available evidence.

What it does

ArchiveWitness is a privacy-first AI Skill designed for UGREEN AI NAS. It helps users transform authorized family photographs and documents into evidence-linked historical narratives in Azerbaijani or English. Rather than inventing missing information, ArchiveWitness distinguishes documented facts, interpretations, and unverified details. Users receive a private draft with source references, unresolved questions, and an opportunity to review everything before saving.

How we built it

I developed the initial ArchiveWitness Skill specification using the official Uliya Skill Development Standard and AI-assisted technical drafting. The workflow coordinates four supported UGOS expert-agent capabilities:

• File Agent: Discovers files within explicitly authorized folders.
• Photo Agent: Examines available image metadata and readable text.
• Document Agent: Extracts information, drafts narratives, and checks supporting evidence.
• Knowledge Agent: Saves approved text records when permission is granted. The package includes SKILL.md and two supporting Markdown references covering privacy, attribution, and validation. The Uliya v0.7.3 static checker returned zero errors and zero warnings.

Challenges we ran into

The greatest design challenge was preventing unsupported historical claims from appearing as verified facts. Other challenges included protecting sensitive family information, maintaining accurate bilingual narratives, restricting access to authorized folders, and handling missing or contradictory evidence. The workflow therefore requires traceable source references and explicit human approval before saving. Actual NAS execution and document-processing reliability remain untested.

Accomplishments that we're proud of

Created an original, evidence-centered family heritage Skill concept.

• Designed a privacy-conscious workflow using registered UGOS expert agents.
• Added explicit source attribution and uncertainty handling.
• Planned Azerbaijani and English narrative generation.
• Implemented approval requirements in the Skill instructions.
• Passed the official local static validation with no reported errors or warnings. These achievements concern the initial Skill specification, not a completed NAS deployment.

What we learned

Responsible AI storytelling requires more than generating fluent text. Historical accuracy depends on provenance, transparent uncertainty, appropriate permissions, and meaningful human review. We also learned that a NAS-based AI workflow can be designed around authorized local resources rather than unnecessary external data transfers.

What's next for ArchiveWitness

If selected as a finalist, ArchiveWitness will undergo end-to-end testing in the official UGREEN AI NAS environment. The validation plan includes bilingual narratives from synthetic family records, folder-access isolation, permission-controlled saving, and conflicting historical evidence. The ultimate goal is to create a practical, trustworthy tool that helps families preserve their heritage without losing control of their memories or personal information.

Built With

• ai-agents
• chatgpt6
• markdown
• ugos
• uliya-skill-development-standard