Tool System

HelseCLI's power comes from its ability to use tools to interact with your environment. The system is designed to be modular and safe.

The efficacy of HelseCLI is derived from its integrated tool system, which enables secure and modular interaction with local and cloud environments.

Functional Tool Classifications

1. File System Operations (file_ops)

Enables direct management of project-level assets and source code.

  • read_file: Telemetry for file content inspection.
  • write_file: Initial creation or non-destructive overwriting of file assets.
  • replace_file_content: Targeted modifications of specific code blocks.
  • list_dir: Recursive discovery of directory structures.

2. Computational Execution Modules (sandbox)

Facilitates the secure execution of logic and retrieval of standard output.

  • execute_python: Runs Python-based logic in a defined sandbox.
  • run_terminal_command: Executes shell operations (e.g., Git, package managers).
  • Runtime Environments: Supports Local (host machine), Docker (containerized), or E2B (cloud-managed) isolation levels.

3. Data Retrieval and Web Integration (web)

Interfaces between the local workspace and external web data sources.

  • web_search: Multi-provider search abstraction (Tavily, Brave, Serper).
  • read_url: Structured Markdown extraction from remote URLs.
  • browser_query: Headless browser automation for deep data retrieval.

4. Advanced Scientific Research Modules (research)

Engineered specifically for data-intensive research and scientific analysis.

  • csv_toolkit: High-performance processing of large-scale datasets.
  • plot_data: Automated generation of technical charts and visualizations.
  • literature_search: Targeted querying of academic and medical databases (e.g., PubMed).

Tool Authorization Protocol

HelseCLI operates under a strict authorization manifest:

  • Standard Mode (Interactive): The AI assistant must request explicit user authorization before initiating any "Write" or "Execute" operation. User may Approve (y) or Refuse (n).
  • Autonomous Agent Mode (!a): Grants the system ephemeral authorization to orchestrate multiple tools autonomously to achieve a higher-level objective.

Security and Isolation Standards

HelseCLI prioritizes computational security. For workflows involving unverified third-party libraries or complex data transformations, the Docker Sandbox is the recommended baseline.

  1. Docker Containerization: Logic executes within a restricted container environment, isolated from the host OS security context.
  2. E2B Managed Sandboxing: Logic executes on a remote, ephemeral virtual machine, providing complete physical isolation from the host system.

Extension and Custom Tool Development

The HelseCLI registry is designed for modular extension. Developers may integrate custom capabilities by:

  1. Implementing a Python class inheriting from the BaseTool abstraction in src/tools/.
  2. Defining the JSON-schema for functional parameters and descriptions.
  3. Overriding the execute() method with custom logic.
  4. Registering the implementation within the src/tools/registry.py manifest.

For detailed development protocols, consult the Custom Tool Development Guide.