Skip to content

solofberlin/hivemind

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

688 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

                          ╭─────────╮
                          │  SOLUM  │
                          ╰─────────╯

The bedrock for living AI agents.

Persistent autonomous agents built on Elixir/OTP.

License: MIT Elixir OTP

Quick Start · What Works Today · Coming Soon · Docs · Discord


Alpha Preview

Demo (GIF) Placeholder

TODO: Replace with real demo GIF.

Suggested path: docs/media/demo-alpha.gif

Solum alpha demo placeholder

Dashboard Screenshot Placeholder

TODO: Replace with current dashboard screenshot.

Suggested path: docs/media/dashboard-alpha.png

Solum dashboard placeholder

Architecture Diagram Placeholder

TODO: Replace with polished architecture diagram.

Suggested path: docs/media/architecture-alpha.png

Solum architecture placeholder


Why Solum

Most agent frameworks run request-by-request: call model, return text, lose continuity.

Solum treats an agent as a long-running OTP process:

  • persistent state
  • supervised fault tolerance
  • async tool execution
  • memory compaction over time
  • sub-agent orchestration via message passing

This follows the architecture direction documented in docs/ARCHITECTURE.md and the Context HUD framing from docs/blog-hivemind-intro.md.


What Works Today (Alpha)

  • Always-on agent processes (GenServer-based) with supervised restarts
  • Multi-agent orchestration (spawn and coordinate sub-agents)
  • Tool execution pipeline with async/background execution
  • Sedimentary memory model and context assembly
  • WAL-backed persistence foundations
  • CLI workflows for running and chatting with agents
  • Live dashboard foundations via Phoenix LiveView
  • Core test suite (500+ tests referenced in project blog)

Coming Soon

  • Polished Context HUD experience end-to-end
  • Better onboarding and first-run setup flow
  • Richer channel/file attachment support across integrations
  • Production release packaging and installer flow
  • Expanded reliability and dashboard UX polish

Architecture (Current Shape)

Solum.Application
│
├── AgentSupervisor          DynamicSupervisor for agent processes
│   ├── Agent "sol"          conversation loop, memory, tool dispatch
│   ├── Agent "coder-*"      sub-agents for specialized tasks
│   └── Agent "..."          horizontally scalable agent pool
│
├── Context.Store            SQLite + compaction pipeline
│   ├── L1 (10-min)
│   ├── L2 (hourly)
│   └── L3-L5 (daily+)
│
├── Tool.Service             async tools + orchestration hooks
├── ChannelSupervisor        Signal / Telegram / CLI adapters
├── Config                   live-reloadable runtime config
└── SolumWeb                 dashboard (Phoenix LiveView)

For deeper detail, see docs/ARCHITECTURE.md.


Quick Start

# Prereq: Elixir 1.19+ and OTP 27+
git clone https://github.com/solum-org/solum.git
cd solum
mix deps.get

# Configure local agent settings
cp config/agent.example.toml config/agent.toml
$EDITOR config/agent.toml

# Setup + run
mix solum.setup
mix solum.start

Open another shell and chat:

mix solum.cli chat --agent sol "Hello."

Dashboard:

open http://localhost:4000

Alpha Call To Action

Solum is in active alpha. If you want to shape the architecture with us:

  • try it on a real workflow
  • file issues with logs and repro steps
  • join design/implementation discussion on Discord

Start here:


 ░▒▓ solum — the ground beneath the agent ▓▒░

MIT License

About

Agent orchestration on the BEAM — Elixir/OTP framework for living AI agents

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages