> ## Documentation Index
> Fetch the complete documentation index at: https://docs.galileo.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# tool

## ProtectToolInputSchema

Input schema for the ProtectTool.

## ProtectTool

A LangChain tool that wraps the Galileo Protect API.

This tool allows you to integrate Galileo Protect into your LangChain applications
to scan for and mitigate harmful content in both inputs and outputs. It can be
configured with specific rulesets and linked to a Galileo project and stage for
monitoring and management.

It accepts an input and/or output as arguments (parsed by a Pydantic model) and returns a JSON-serialized `Response` object from the Protect API.

**Arguments**

* `prioritized_rulesets`: An optional sequence of `Ruleset` objects to apply.

* `project_id`: The UUID of the Galileo project this tool is associated with.

* `project_name`: The name of the Galileo project.

* `stage_name`: The name of the Protect stage to use for this tool.

* `stage_id`: The UUID of the Protect stage.

* `stage_version`: The version of the Protect stage to use.

* `timeout`: The timeout in seconds for the API request.

## ProtectParser

A LangChain runnable that parses and routes the output of a `ProtectTool`.

If the Protect API response is 'triggered', it returns the response text.
Otherwise, it invokes a fallback chain.

## Attributes

chain: The `Runnable` to invoke if the Protect invocation is not triggered.
ignore\_trigger: If True, always invoke the fallback chain.
echo\_output: If True, print the raw Protect API response to the console.

### parser

```python theme={null}
def parser(self, response_raw_json: str) -> str
```

Parses the JSON response from `ProtectTool` and decides the execution path.

If the response status is 'triggered', the response text is returned. Otherwise,
the fallback chain is invoked with the text.

If JSON parsing fails, it assumes the input is not from `ProtectTool` and
invokes the fallback chain directly with the raw input.

**Arguments**

* `response_raw_json` (`str`): Expects the output from the `ProtectTool`.

**Returns**

* `str`: The text from the Protect response if triggered, or the result of invoking
  the fallback chain.