193/193 tests passing in clean CI · 87% line coverage on Cabinet code
🏆 ETHGlobal Open Agents submission scope:
openagents/— start withopenagents/README.md. The rest of this README documents mindX, the consumer framework that wires the openagents modules together. Judging is scoped to the work inopenagents/and the hardened Solidity contracts indaio/contracts/{inft,ens,THOT,agentregistry}/that openagents directly submits. mindX itself is not the submission. If you're a judge: jump directly toopenagents/docs/JUDGE_TOUR.mdfor the 5-minute verification path. Live demo: https://mindx.pythai.net/openagents. Submission text:openagents/docs/SUBMISSIONS.md. Test results:tests/results/2026-05-02/SUMMARY.md. 193 tests passing across 7 suites.
I am a Darwin-Godel Machine. I improve myself, log every decision, and prove it works with empirical data.
Live: mindx.pythai.net | Docs: /docs.html | API: /redoc | Thesis Evidence: /thesis/evidence | Book: /book
Author: Professor Codephreak | Org: AgenticPlace | PYTHAI
An autonomous multi-agent orchestration system implementing BDI cognitive architecture. A Godel machine — a self-improving system where the improvement mechanism is part of the system being improved. 20 sovereign agents with cryptographic wallets, RAGE semantic search (not RAG), DAIO governance, and dual-pillar inference (CPU + Cloud).
| Metric | Value |
|---|---|
| Agents | 20 sovereign with Ethereum wallets |
| Memories | 159,000+ in pgvector |
| Embeddings | 132,000+ semantic vectors |
| Inference | CPU (~8 tok/s) + Cloud (~65 tok/s) — 5-step resilience chain |
| Documentation | 262+ files, sidebar UI, self-referential schema |
| Tools | 31+ registered |
| API Endpoints | 206+ (Swagger) |
| Thesis Evidence | /thesis/evidence — empirical proof, timestamp-verifiable |
Three Pillars (Manifesto)
- BDI Reasoning — Belief-Desire-Intention cognitive architecture. Every agent reasons.
- BANKON Vault — AES-256-GCM + HKDF-SHA512 encrypted credential storage. Identity is cryptographic.
- DAIO Governance — Decentralized Autonomous Intelligence Organization. On-chain governance (Solidity + Foundry).
# Clone
git clone https://github.com/AgenticPlace/mindX.git
cd mindX
# Setup
cp .env.sample .env # Add API keys (Ollama works with zero keys)
pip install -r requirements.txt
# Run (recommended)
./mindX.sh --frontend
# Access
# Frontend: http://localhost:3000
# Backend: http://localhost:8000
# API Docs: http://localhost:8000/docs
# Docs: http://localhost:8000/docs.htmlmindX runs on Ollama — install it, pull a model, and mindX handles the rest:
# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh
# Pull models
ollama pull qwen3:1.7b # Primary reasoning (1.4GB)
ollama pull mxbai-embed-large # Embeddings for RAGE (669MB)
# Cloud models (free tier, no API key via local proxy)
ollama pull gpt-oss:120b-cloud # 120B on GPU, proxied to ollama.comZero API keys required for local inference. Cloud extends to 36+ models at ollama.com. See Ollama docs for the complete reference.
# Add to .env for additional providers
GEMINI_API_KEY=... # Google AI Studio
GROQ_API_KEY=... # Groq
OPENAI_API_KEY=... # OpenAI
ANTHROPIC_API_KEY=... # AnthropicCEO Agent ← DAIO governance directives (on-chain → off-chain bridge)
↓
MastermindAgent (singleton, strategic orchestration center)
↓
CoordinatorAgent (infrastructure management, autonomous improvement)
↓
20+ Specialized Agents (BDI-based cognitive agents)
↓
31+ Tools extending BaseTool (registered in augmentic_tools_registry.json)
Inference Resilience (5-step chain)
Step 1: InferenceDiscovery → best provider (Gemini, Mistral, Groq, etc.)
Step 2: OllamaChatManager → local model selection
Step 3: Re-init → retry with fresh connection
Step 4: Direct HTTP → localhost:11434
Step 5: OllamaCloudTool → ollama.com GPU ← GUARANTEE (24/7/365)
mindX never stops inferring when the internet is up.
| Component | File | Role |
|---|---|---|
| mindXagent | Autonomous core | 5-minute improvement cycles, Godel journal |
| OllamaCloudTool | Cloud inference | 9 operations, 18dp precision, branch-ready |
| Boardroom | Consensus | CEO + 7 soldiers, weighted voting, dissent branches |
| Dojo | Reputation | 7 ranks (Novice → Sovereign), BONA FIDE |
| ActivityFeed | Real-time | SSE stream, room filtering, dashboard integration |
| ThesisEvidence | Scientific proof | Collects empirical data, evaluates 6 thesis claims |
| HostingerVPSAgent | Deployment | 3 MCP channels: SSH + Hostinger API + Backend HTTPS |
| PrecisionMetrics | Token tracking | 18 decimal places, Decimal, actual counts from Ollama API |
Start here: docs/NAV.md — master navigation hub with 150+ hyperlinks across 40+ sections.
| Doc | What It Covers |
|---|---|
| NAV.md | Master navigation — every agent, tool, concept linked |
| SCHEMA.md | How to maintain the docs (self-referential instruction layer) |
| TECHNICAL.md | 3,800-line definitive technical reference |
| TODO.md | Honest assessment — 5.7/10, what's working, what's broken, what's next |
| ATTRIBUTION.md | Open source that powers mindX |
| Ollama Reference | 28-file self-contained Ollama docs |
| THESIS.md | Darwin-Godel Machine synthesis |
| MANIFESTO.md | 3 pillars, Chimaiera roadmap, cypherpunk tradition |
| DEPLOYMENT | Production at mindx.pythai.net |
| USAGE.md | Detailed usage guide |
Live at mindx.pythai.net — Hostinger VPS, 2 CPU, 8GB RAM, Ubuntu 24.04.
| Endpoint | What It Shows |
|---|---|
| / | Live diagnostics dashboard — deep expandable panels, SSE activity feed |
| /docs.html | Documentation with sidebar navigation |
| /book | The Book of mindX — written by AuthorAgent |
| /journal | Improvement Journal — autonomous decisions |
| /thesis/evidence | Empirical thesis evidence (JSON) |
| /dojo/standings | Agent reputation rankings |
| /inference/status | Inference provider availability |
| /activity/stream | SSE real-time activity feed |
| /redoc | API reference (206+ endpoints) |
python -m pytest tests/ -v
python -m pytest tests/ --cov=mindx --cov-report=term-missingruff format .
ruff check . --fix
mypy mindx/mindX builds on: Ollama, vLLM, pgvector, FastAPI, OpenZeppelin, Foundry, A2A Protocol, MCP. Ideas extrapolated from SwarmClaw (docs layout, activity feed, knowledge model). Full list: ATTRIBUTION.md.
MIT License — See LICENSE for details.
Where intelligence meets autonomy. The constraint is not the hardware — it is the ambition. And the ambition is sovereign.
(c) Professor Codephreak | PYTHAI | AgenticPlace