Becasue I’m working in Health Informatics at the moment, initiatives like OpenMed are really interesting to me. I’ve gotten it working on my Raspberry Pi.
1. Prepare the Directory
Ensure you are in the parent folder where you want the project to live. If the folder OpenMed already exists and has failed build files in it, it’s safest to start fresh.
mkdir -p ~/Development/OpenMed
cd ~/Development/OpenMed
2. Clone the Correct Repository
Use the official repository URL. By adding the . at the end, you clone the files directly into your current folder instead of creating a nested OpenMed/openmed structure.
git clone https://github.com/maziyarpanahi/openmed.git .
3. Environment Setup (Recommended)
On a Raspberry Pi, it is best practice to use a virtual environment to avoid conflicts with system packages. If you are using uv (as recommended by the OpenMed docs):
# Create and activate a virtual environment
uv venv
source .venv/bin/activate
4. Install Dependencies
This is where the previous error occurred. Now that pyproject.toml is present in your current directory, the installer will recognize it as a project.
Important: Use the [hf] flag. The [mlx] flag in the docs is exclusively for Apple Silicon Macs and will fail on a Raspberry Pi. However if you’re running this on Apple Silicon, you’ll get increased performance with the mlx flag.
# Install the project in editable mode with Hugging Face dependencies
uv pip install -e ".[hf]"
Note: If you aren’t using uv, simply use pip install -e “.[hf]”.
5. Verify the Installation
To make sure everything is linked correctly and the CLI is responsive, run:
openmed --help
6. Install the OpenMed Agent
# Ensure your venv is still active (you see (.venv))
uv pip install openmed-agent
then
(openmed) username@blue-meanie:~/Development/OpenMed/openmed $ openmed --help
usage: openmed [-h] [--version] [--config-path CONFIG_PATH]
[--agent {clinical,plan,default}] [--mcp-enabled]
{tui,agent,config,login,oauth,codex,mcp,update} ...
OpenMed Agent — AI-powered medical assistant with TUI.
positional arguments:
{tui,agent,config,login,oauth,codex,mcp,update}
tui Launch interactive terminal UI for clinical NER analysis.
agent Launch OpenMed Agent TUI - AI-powered medical assistant.
config Inspect or modify OpenMed CLI configuration.
login Authenticate for Codex via OAuth.
oauth Manage Codex OAuth authentication.
codex Codex authentication and management.
mcp Manage remote MCP server connectors.
update Update OpenMed to the latest version.
options:
-h, --help show this help message and exit
--version show program's version number and exit
--config-path CONFIG_PATH
Override the configuration file path.
--agent {clinical,plan,default}
Agent mode for the TUI (clinical or plan).
--mcp-enabled Enable all supported remote MCP servers before running the
command. This persists in MCP config.
Cool. Almost there. this will work with the REST but you might tant the Terminal User Interface
uv pip install openmed-agent[tui]
Then finally
openmed tui
Now….it seem by default it’s using OpenAI….and I want it to run local only. Which just means doing more investigation.