nimo-controller 0.0.1

NIMO Controller: MCP-based interface for self-driving laboratories

nimo-controller

Installation

Installation requires uv.

git clone https://github.com/NIMS-DA/nimo-controller.git
cd nimo-controller
uv sync

Examples

Bayesian optimization

The example_mcp/ackley directory contains an example MCP server that demonstrates how to use PHYSBO with the Ackley function.

Start the MCP server:

cd example_mcp/ackley
uv run mcp_ackley.py

In a separate terminal, launch NIMO Controller:

cd nimo-controller
uv run nimo-controller

Then open http://127.0.0.1:8888/ in your browser. At this point you will see a warning indicating that no candidates file has been loaded yet.

Click the Upload candidates file button and upload example_mcp/ackley/candidates.csv. Once uploaded, the x1 and x2 blocks will appear in the NIMO toolbox.

Use the blocks to build the workflow shown below:

Press the Run button to execute the workflow.

Note that the target is the negated Ackley function, since PHYSBO performs maximization.

Phase diagram construction

The example_mcp/phase_diagram directory contains an example MCP server that demonstrates the PDC algorithm by constructing a phase diagram for the fictional material X described in the IvoryOS integration guide.

Start the MCP server:

cd example_mcp/phase_diagram
uv run mcp_phase_diagram.py

Restart NIMO Controller, then upload example_mcp/phase_diagram/candidates.csv via the Upload candidates file button.

Build the workflow shown below. After running the workflow, a phase diagram will be constructed.

Configuration

Configuration is loaded from the following sources, in order of precedence:

1. The path specified by --config <path>
2. ./config.yaml in the current directory
3. <user-config-dir>/nimo-controller/config.yaml (for example, ~/.config/nimo-controller/config.yaml)
4. The default configuration bundled with the package

Example config.yaml:

model: "gpt-4o-mini"
port: 8888

# Optional: route requests to an OpenAI-compatible local endpoint (e.g. Ollama).
# openai_base_url: "http://127.0.0.1:11434/v1"
# openai_api_key: "ollama"

mcp_servers:
  sdl: "http://127.0.0.1:8001/mcp"

Runtime data, including the uploaded candidates.csv and optimization results, is written to <user-data-dir>/nimo-controller/ (for example, ~/.local/share/nimo-controller/).