Danielius Studio

AI agents & automation

AI Legacy Preservation Platform

Upload documents, voice, and video. AI answers questions about you from what you uploaded.

Architecture: Upload & Indexing

User the person preserving their legacy Upload Surface web app, login-gated Documents PDF, Word, text Voice Recording spoken stories, memories Video recorded interviews, clips Text Extraction parses PDF, Word, plain text Storage per-user folders, raw files Whisper Transcription audio and video to text Privacy Layer per-user isolation, row rules Chunking & Tagging small chunks, user ID on each Backups nightly to object storage Embedding OpenAI text-embedding-3-small Vector Store Postgres + pgvector signs in uploads docs voice uploads video transcribes all text tags per user nightly backup chunks fingerprints stored

Architecture: AI Q&A Pipeline

User or a family member with access Chat UI asks a question, embeds it Vector Store Postgres + pgvector Retrieval top matching chunks, scoped to user RAG Composer context + question + system prompt LLM GPT-4o-mini or Claude Haiku Citations source uploads, clickable Answer text response, grounded Voice Output ElevenLabs reads the answer asks search similarity matched chunks context + question source-tagged ElevenLabs TTS

Want this built?

I architect, build, and ship engagements like this one.