Skip to content
Satya Kishore

Projects

Case study

OpenHuman

Local-first personal AI with memory and agent orchestration — studied from existing open-source work, not claimed as original.

Fork / Based on existing workBased on existing open-source work

Local AI · Agents

01 / Overview

Local-first personal AI / memory-oriented project based on existing open-source work (a fork of tinyhumansai/openhuman).

  • Rust
  • TypeScript
  • SQLite
  • Ollama
  • Tauri

02 / Problem

Personal AI that remembers, orchestrates agents, and stays local-first is a dense systems problem. The fastest way to learn that shape is to run and study a serious existing codebase rather than re-inventing one from slogans.

03 / Approach

This portfolio entry is a fork of OpenHuman (tinyhumansai/openhuman), used to study local-first memory, agent fleets, and workflows. It is not presented as an original product.

04 / Architecture

  1. 01

    Ingest

    Integrations / local sources

  2. 02

    Memory tree

    SQLite + Markdown

  3. 03

    Orchestration

    Agent graphs

  4. 04

    Workflows

    Reviewed automations

  5. 05

    Local / routed models

The pipeline above is a high-level reading of the upstream OpenHuman project. It is not a claim that this architecture was designed here. Changes relative to upstream are not documented in the fork.

05 / Implementation

  • Repository is a fork of tinyhumansai/openhuman (Rust core, Node/pnpm workspace, desktop shell).
  • Upstream describes a SQLite-backed memory tree mirrored as Markdown, plus agent orchestration and workflow tooling.
  • No separate original implementation log is published on this fork.

06 / Engineering decisions

Study the real system
Local-first memory and multi-agent orchestration are easier to understand from a running open-source project than from a from-scratch toy.
Do not re-attribute upstream work
Feature lists, architecture, and performance claims belong to the upstream project unless a specific local change is documented. None is documented here.

07 / Evaluation

No evaluation is claimed for this fork. Upstream marketing metrics are not repeated as personal results.

08 / Limitations

  • Based on existing open-source work — not original authorship.
  • Differences from upstream (if any) have not been documented and should be treated as unverified.
  • Do not read this card as a shipped personal-AI product.

09 / Repository

github.com/Kishore-3021/openhuman

No public demo listed.