Embedded

#What it does

Embedded turns voice memos into searchable, summarized knowledge. Record a thought, and AI handles the rest — transcription, speaker identification, executive summaries, and semantic search across everything you’ve ever said.

#The AI Pipeline

Every voice memo flows through a multi-stage cloud pipeline:

• Transcription — OpenAI Whisper converts speech to text with high accuracy
• Summarization — GPT-4o-mini generates concise executive summaries
• Speaker ID — Pyannote.ai identifies who said what in multi-person recordings
• Embeddings — Google Gemini creates 3072-dimensional vectors for semantic search
• Chunking — Long recordings are automatically split into overlapping chunks for unlimited file sizes

#Architecture

• iOS App — Native Swift/SwiftUI with Firebase Auth and Storage
• Backend — Google Cloud Functions (Python) triggered by audio uploads
• Database — Firestore for metadata, Supabase + pgvector for embeddings
• Security — Row Level Security ensures users only access their own data

#Embedded Connect

The open-source connector ecosystem lets you export your data to the tools you already use:

• Obsidian — Smart Vault with dashboards, people routing, relationship graphs, and 1:1 file management
• Notion — Database pages with structured properties and rich content
• JSON Export — Full data dump with optional embedding vectors
• Python SDK — Authenticate and fetch your memos in 3 lines of code

View Embedded Connect on GitHub

#Stack