ROS Package bob_central

This package is a General Central Orchestration Brain-Mesh System designed for building and hosting self-evolving, autonomous AI entities within isolated container environments. It represents an AI deeply integrated into a ROS 2 environment, leveraging the full power of the ROS 2 ecosystem (topics, services, and parameters) for real-world interaction and self-monitoring.

The Central Nervous System of the Bob ROS Ecosystem.

bob_central provides the essential infrastructure for orchestrating complex, multi-modular AI agents in a ROS 2 environment. It handles everything from high-level decision making (Orchestrator) to real-time system visualization (nviz), stateful code engineering (REPL), and autonomous documentation management (Knowledge Graph).

Core Concept

At its heart, bob_central manages a “Brain-Mesh” of interconnected specialized nodes. The system is not monolithic; it is a distributed network of intelligence where every component is replaceable and extensible.

Key Features

• Recursive Reasoning (RLM Core): Multi-step internal dialogue using expert personas (Architect, Critic, Planner, Debugger) to decompose complex tasks.

• Persistent Python REPL: A stateful engineering environment for iterative code development and system manipulation, preserving state across sessions.

• Centralized Orchestration: A powerful node that manages conversation flows, busy-locking, and tool calls.

• Visual Telemetry (nviz): High-performance, event-driven dashboard rendering (8-bit grayscale bitmaps) with real-time status indicators.

• Autonomous Knowledge Graph: On-demand technical documentation fetching and indexing for AI context.

• Self-Evolution Framework: Pure ROS 2 native infrastructure for agents to modify and expand their own capabilities.

Recursive Thought (RLM)

The Recursive Language Model core enables Eva to use the perform_thought tool to consult internal specialists before executing sensitive actions.

Persistent Engineering (REPL)

The repl_kernel skill provides Eva with a permanent engineering workspace.

• Persistent State: Variables, imports, and function definitions persist as long as the stack is running.

• Safety: Isolated execution via a dedicated repl_node with 15s timeouts and capture of all stdout/stderr output.

Ecosystem Management

The Docker Ecosystem

To manage the complex set of services, a master management script is provided in the docker/ directory.

Quick Management:

./docker/manage.sh up      # Start the entire ecosystem
./docker/manage.sh down    # Stop all services
./docker/manage.sh build   # Rebuild local images

Compose Stacks

File	Description
compose-base.yaml	Core logic (eva-base) and API Gateway (eva-api-gate).
compose-nviz.yaml	Visual dashboard streamer (eva-nviz-streamer).
compose-tti.yaml	Image generation engine (eva-artist).
compose-tbot.yaml	Twitch Chatbot & Twitch Integration stack.
compose-dashboard.yaml	Dedicated telemetry and dashboard automation.
compose-q3tts.yaml	Text-to-Speech engine (Qwen3-TTS).
compose-gitea.yaml	Local Git infrastructure and CI runner.
compose-inference.yaml	LLM inference servers (Vision/Reasoning).
compose-qdrant.yaml	Vector database for long-term memory.
compose.face.yaml	Facial animation and sentiment visualization engine.

Security Features

ROS 2 API

Nodes & Topics

Topic	Type	Description
/eva/user_query	std_msgs/String	Universal input channel for user queries.
/eva/repl/input	std_msgs/String	Raw Python code feed for the persistent REPL node.
/eva/repl/output	std_msgs/String	Captured output from the engineering workspace.
/eva/dashboard/visual_trigger	std_msgs/String	Internal status triggers (busy/idle/thinking) for UI.
/eva/llm_stream	std_msgs/String	Real-time token stream for low-latency interfaces.

Development & Evolution

• Linter Compliant: 100% compliance with ament_lint_auto, flake8, and pep257.

• Standardized Skills: All tools are documented via SKILL.md using the Anthropic Agent Skill standard.

• Extensible Architecture: Designed for autonomous self-evolution.

Snapshots

Current architectural state visualized as ROS graph diagrams.

ROS RQT Dynamic Reconfigure GUI’s