Configuration & Endpoints

Minimal settings to run locally or in Docker. Values shown reflect defaults in the image and application.yaml.

Environment Variables

• OPENAI_API_KEY — API key for OpenAI (optional if using other providers)
• GEMINI_API_KEY — API key for Gemini (optional)
• ANTHROPIC_API_KEY — API key for Anthropic (optional)
• OTEL_EXPORTER_OTLP_ENDPOINT — OTLP collector endpoint (default: http://localhost:4317)
• TS_RUNTIME_URL — TypeScript runtime URL (default: http://localhost:7070) — optional/placeholder
• FOUNDATION_DIR — Foundation directory (default: /var/foundation)
• INFERENCE_DEFAULT_PROVIDER — AI provider (openai, gemini, or anthropic)

Process Arguments

• —process=validate — validates foundation data and exits
• —process=regression — runs regression test suite and exits
• —process=standard (default) — starts MCP server + orchestrator

These can be passed via APP_ARGS in Docker:

# Validation mode
docker run --rm -e APP_ARGS="--process=validate" admingentoro/gentoro:latest
 
# Regression mode
docker run --rm -e APP_ARGS="--process=regression" admingentoro/gentoro:latest
 
# Standard mode (default)
docker run --rm -p 8080:8080 admingentoro/gentoro:latest

Foundation Directory Structure

The process modes expect a foundation directory with the following structure:

foundation/
├── Agent.md                    # Required: Agent configuration
├── apis/                       # Optional: OpenAPI specifications
│   └── *.yaml
├── docs/                       # Optional: Documentation files
│   └── *.md
├── regression/                 # Optional: Regression test files
│   └── *.yaml
└── state/                      # Auto-generated: Knowledge base state
    └── knowledge-base-state.json  # Indexed docs, APIs, and content signature

MCP Endpoint

• Path: /mcp (HTTP stream)
• Example URL: http://localhost:8080/mcp

Client connections must speak the Model Context Protocol over HTTP streaming.