Tools¶

OSA provides a modular tool system for document retrieval, validation, and knowledge search. Tools are organized by domain and can be composed into assistant workflows.

Modular Design¶

Each assistant in OSA has its own set of tools that can be independently developed and added. When a new assistant is added to OSA, its documentation is automatically populated here.

Assistant Tool Sets¶

HED Tools¶

Tools for working with Hierarchical Event Descriptors (HED).

• HED Tools - Document retrieval, validation, tag suggestions

Status: Available

BIDS Tools¶

Tools for working with Brain Imaging Data Structure (BIDS).

Status: Planned - will be added when BIDS assistant is implemented

EEGLAB Tools¶

Tools for working with EEGLAB analysis toolbox.

Status: Planned - will be added when EEGLAB assistant is implemented

Common Tool Categories¶

Each assistant typically provides tools in these categories:

Document Retrieval¶

Fetch documentation from official sources:

Validation¶

Validate domain-specific data:

Knowledge Search¶

Search community knowledge:

Tool Architecture¶

Tools follow a common pattern using LangChain:

from langchain_core.tools import tool
from pydantic import BaseModel, Field

class RetrieveDocsInput(BaseModel):
    """Input for document retrieval."""
    topic: str = Field(description="Topic to search for")
    max_docs: int = Field(default=3, description="Maximum documents")

@tool(args_schema=RetrieveDocsInput)
def retrieve_docs(topic: str, max_docs: int = 3) -> str:
    """Retrieve documentation for a topic."""
    # Implementation
    pass

Tool Permissions¶

Some tools require explicit user permission:

Adding Custom Tools¶

Communities can add their own tools:

1. Create a tool function with @tool decorator
2. Add to the tool registry in src/tools/
3. Include in assistant configuration
4. Add documentation in docs/osa/tools/{domain}.md

See Development for detailed instructions.