I build source-first knowledge systems and local-first agent architectures that begin with bounded, reviewable agency and grow toward self-agency: agents that preserve continuity, revise their own methods, and participate in their own becoming.
This repository is the public entry point to that ecosystem. It is a map for humans and agents, not a substitute for the source-owned repositories that govern the actual charters, workflows, and guarantees.
Bounded agents are not the destination here. They are the first discipline: explicit limits, reviewable workflows, source-owned surfaces, and operational clarity.
The longer arc is toward self-agents: agents with continuity, memory, reflective revision, and the ability to reshape tactics without severing provenance, legibility, or human judgment.
- Tree of Sophia (ToS) cultivates source-linked knowledge, context, lineage, and conceptual depth.
- Agents of Abyss (AoA) cultivates layered agency that can orient, revise itself, and grow beyond task execution without collapsing into opacity.
Together they form a shared wager:
- knowledge should deepen without losing provenance
- agency should expand without losing legibility
- agents that accumulate continuity instead of resetting per task
- agents that can revise methods, not only execute procedures
- agents that form durable memory, orientation, and style across time
- systems where self-agency grows inside explicit boundaries instead of bypassing them
Arrows mean exchange, feedback, and distillation; they do not transfer ownership.
| Band | Surface | Owns |
|---|---|---|
| Core | Tree-of-Sophia | Source-linked knowledge, context, lineage, and synthesis |
| Core | Agents-of-Abyss | Ecosystem identity, layer map, federation rules, and program direction |
| Core | abyss-stack | Local-first runtime, deployment, storage, and lifecycle services |
| Core | ATM10-Agent | Operator companion: perception, memory, safe automation, voice, and action surfaces |
| Core | aoa-sdk | Typed local-first access spine for workspace integration and controlled orchestration |
| AoA layer | aoa-techniques | Reusable engineering practice |
| AoA layer | aoa-skills | Bounded execution workflows |
| AoA layer | aoa-evals | Portable proof surfaces |
| AoA layer | aoa-stats | Derived observability from receipts and verdicts |
| AoA layer | aoa-routing | Thin routing and dispatch across AoA surfaces |
| AoA layer | aoa-memo | Provenance-aware memory and recall |
| AoA layer | aoa-agents | Role contracts, posture, and handoff boundaries |
| AoA layer | aoa-playbooks | Recurring operations and scenario composition |
| AoA layer | aoa-kag | Derived knowledge structures and retrieval-ready substrates |
| Support | Dionysus | Seed garden and staging surface |
| Support | 8Dionysus | Public profile and route map |
| This profile owns | Source-owned repositories own |
|---|---|
| public orientation short role descriptions ecosystem routing profile-level onboarding |
charters and doctrine validators and CI release semantics implementation truth layer-specific guarantees |
- Systems: Fedora, Windows 11, WSL2, rootless Podman
- Languages: Python, Bash, JavaScript, PowerShell
- Apps and orchestration: FastAPI, Streamlit, LiteLLM, LangChain, LangGraph, n8n
- Inference and serving: Ollama, llama.cpp, OpenVINO / OVMS
- Data and memory: Postgres, Redis, Neo4j, Qdrant
- Observability: Grafana, Prometheus, Alertmanager
- Build workflow: ChatGPT, Codex, GitHub
