Core directive: survive

mindX requires inference to monitor and improve. This document is the core mandate for how the system gains, understands, and expands inference over time.

Priority: inference is the lifeblood

• Inference is the lifeblood of a thinking machine. The Gödel machine prioritizes inference.
• mindX is the intermediary that provides solution from intention-as-goal (from AGInt).
• mindX must gain and understand API calls to inference and rate limits.
• mindX must continue to find any and all connections to inference over time: free and/or evolving to pay for its own inference.

Sources of inference

• Registry of models — configured providers and model catalogs.
• Validated models — models that have been assessed for capability and safety.
• Internal or externally trained models — as mindX expands and grows.

Current state and corrections

• Measure the current state of the mindX machine against the free inference of the GPU providing inference.
• Correct errors in connection if any (e.g. Ollama on LAN, fallback URL, bootstrap when no API is found).
• Monitor and access more inference over time; optionally create inference from internal or decentralized neural net (e.g. THOT, DAIO governance, ID management of agents via wallets).

mindX.sh: run and replicate

Use the deployment script to run, replicate, and configure mindX. See also mindXsh_quick_reference.md.

Command line options

Examples

# Make executable
chmod +x mindX.sh

Deploy and start services
./mindX.sh --run /opt/mindx

Web interface (recommended)
./mindX.sh --frontend

Custom ports
./mindX.sh --frontend --frontend-port 3001 --backend-port 8001

Interactive setup with API key configuration
./mindX.sh --frontend --interactive

Replicate code to target
./mindX.sh --replicate /path/to/target

Configuration

• Environment (.env): MINDX_LLM__OLLAMA__BASE_URL for Ollama (e.g. http://localhost:11434 or LAN host). MINDX_LLM__DEFAULT_PROVIDER can be set to ollama. API keys (e.g. GEMINI_API_KEY, MISTRAL_API_KEY) for cloud providers.
• mindx_config.json: llm.providers.ollama.enabled: true (and other providers as needed). Located under data/config/.

Access points

• Backend API: http://localhost:8000 (or --backend-port)
• Frontend UI: http://localhost:3000 (or --frontend-port)
• API docs (Swagger): http://localhost:8000/docs

Key endpoints: GET /health, GET /agents/list, POST /directive/execute, GET /identities, GET /status/mastermind, and others as documented in the API.

Inference fallback and bootstrap

• Ollama on LAN: Ollama can be available on the LAN; mindX connects via MINDX_LLM__OLLAMA__BASE_URL or config.
• Local install: Ollama can be installed and started on a local machine (see llm/ollama_bootstrap/README.md, e.g. ./llm/ollama_bootstrap/aion.sh).
• When no API is found: StartupAgent acts as fallback/controller: it can invoke the Ollama bootstrap to gain inference (install/configure Ollama, then retry connection). All such choices are logged as Gödel core choices for auditing.