MB Logo
Home Work Projects Notes Practice About
Active

LifeOS

AI-powered knowledge system that transforms notes, transcripts, and resources into expert personas, citation-grounded conversations, and autonomous workflows.

Project LifeOS
Status Active
Last Updated June 2026
Key Tags Python, FastAPI, RAG

Live Demo & Source

Demo Expert: Dalai Lama

The public demo includes a synthesized expert persona built from books, talks, and transcripts, allowing visitors to explore how LifeOS transforms source material into conversational expertise.

LifeOS Main Interface

LifeOS transforms notes, transcripts, books, videos, and resources into searchable expert personas and conversational knowledge systems.

Project Outcomes

  • Public interactive demo
  • Expert personas synthesized from source material
  • Citation-grounded conversational retrieval
  • Automated YouTube and web knowledge ingestion
  • MCP-compatible knowledge server
  • Autonomous code review and pull request workflows

Overview

LifeOS is an experiment in turning personal knowledge into an active thinking environment.

Rather than storing notes, transcripts, books, and resources in isolated systems, the platform transforms them into searchable expert personas, structured knowledge libraries, and conversational AI assistants grounded in source material.

The goal is not simply retrieval.

The goal is helping knowledge become more useful.

Why I Built It

Over the years I accumulated notes, transcripts, books, courses, project artifacts, and research across multiple systems.

Finding information was rarely the problem.

Making use of it was.

LifeOS began as an exploration of whether AI could help transform a collection of information into a network of experts, principles, and actionable insights.

The project combines personal knowledge management, retrieval systems, expert synthesis, and AI-assisted workflows into a single environment.

Core Question

Can personal knowledge become an active thinking partner rather than a passive storage system?

Key Components

Expert Personas

LifeOS Expert Personas

LifeOS automatically synthesizes expert profiles, playbooks, and principles from ingested knowledge.

Rather than manually searching documents, users can interact with curated expert perspectives grounded in source material.

Conversational Retrieval

LifeOS Conversational Chat

Multi-turn chat with citations and expert routing.

Questions can be answered by a specific expert persona or across the entire knowledge base.

Knowledge Library

LifeOS Knowledge Library

Structured storage for notes, transcripts, summaries, and source material.

Every resource remains searchable, organized, and linked to its original source.

Autonomous Hermes Loop

Hermes Autonomous Workflow Diagram

LifeOS does not stop at retrieval.

The Hermes system reviews new knowledge, identifies actionable insights, and can propose code improvements through automated pull requests.

System Architecture

LifeOS combines knowledge ingestion, expert synthesis, retrieval, and autonomous workflows into a single local-first architecture.

LifeOS Architecture Diagram

Technology Stack

Core Platform

  • Python
  • Streamlit
  • FastAPI

Knowledge Layer

  • Markdown
  • YAML Frontmatter
  • SQLite FTS5

AI Layer

  • Google Gemini
  • Azure OpenAI
  • OpenRouter

Agent Systems

  • MCP (Model Context Protocol)
  • Hermes Agent
  • GitHub Pull Request Automation

Ingestion

  • yt-dlp
  • BeautifulSoup
  • Jina Reader

Testing & Quality

  • pytest
  • GitHub Actions

Current Status

Active development. Currently deployed as a public demo and open-source project.

Current capabilities include:

  • Expert synthesis
  • Conversational retrieval
  • YouTube ingestion
  • Web ingestion
  • MCP integration
  • Autonomous review workflows

Future work includes offline models, richer knowledge visualizations, and expanded agent capabilities.

Lessons Learned

The hardest challenge was not retrieval.

It was creating structures that make knowledge useful.

Large language models are increasingly capable of synthesizing information, but the quality of their outputs depends heavily on how knowledge is organized.

LifeOS reinforced the idea that knowledge management is less about collecting information and more about creating clarity, context, and action.

← All Projects Markus Berlit