Start with a small workspace

Before you start#

This is a source-build guide. You need the repository, its configured Python environment and a reachable model service. It is not a one-command public installer. Follow the repository setup instructions for your branch before using these entry points.

The Python CLI defaults to a local Ollama endpoint. Inspect your configuration: a remote endpoint changes where project material is processed.

• Use a test project without credentials or production data.
• Keep a version-control checkpoint or backup.
• Confirm the model endpoint and the tools enabled in your environment.

Start the Python CLI#

From the repository root, select a workspace using --workspace. Omitting --prompt starts an interactive session. The path below is a placeholder: replace it with the directory you intend to inspect.

python3 -m agent_system.cli --workspace /path/to/project

Ask for an inspection first#

A read-only question is a useful first task. Ask for an explanation of a focused piece of the project before requesting a change. Check which files the agent actually inspected, rather than relying only on its final summary.

python3 -m agent_system.cli --workspace /path/to/project \
  --prompt "inspect the project and report what you find"

Review before you continue#

An initial run should establish whether the environment works and whether the output is useful. Review the scope and any tool activity. If the model or runtime is unavailable, fix that dependency; the website illustrations are not a fallback response from the model.

For a visual interface, continue with the local runtime guide. Before requesting a write or connecting another service, read permissions and data flow.