feature(client): tool parameters raw input mode, togglegroup component

[jjjuk]

Tool parameters raw input mode, togglegroup ui component

Hi guys! Needed this one for myself, created a PR)

Form/Raw switch for tool parameters input.
Raw mode uses DynamicJsonForm.
Added forceSimpleJson property to DynamicJsonForm to hide it's own form/json switch.
New ui component togglegroup with similar styling as tabs used for switch, might be useful for other features.

Written as simple as possible, in the project's codestyle.

Motivation and Context

In some tools, empty string instead null is required. But input form converts empty strings to nulls.
So the quick solution is to add DynamicJsonForm as the option to edit tool parameters.
This is usefull if you want more contorl over parameters value, bypass schema, use empty strings, etc.

How Has This Been Tested?

Manually, feature is very basic.

Breaking Changes

Nothing

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update

Checklist

• I have read the MCP Documentation
• My code follows the repository's style guidelines
• New and existing tests pass locally
• I have added appropriate error handling
• I have added or updated documentation as needed

Additional context

Need advice for necessary docs and tests for this feature.

[olaservo]

Hi @jjjuk , I want to make sure I understand the reason behind adding these changes. It sounds like you're saying that empty strings aren't being handled correctly in some cases? I'd prefer not to add more complexity to the form unless its really necessary.

[jjjuk]

Hi, @olaservo! I fully understand your conserns, here is my explanation!

I believe that raw mode is a useful feature for debugging tool calls.

For example, Postman has a simillar feature.
Sometimes it's handy to pass something unexpected or to see the exact value of the parameters.

In my case there is no other reasonable way to specify an empty string. An empty textarea without default value in schema returns "". But cognee-mcp__cognify gives a default value of null, which I think is a mistake. So the form applies defult null values.

Here is cognee-mcp__cognify parameters schema:

{
  "type": "object",
  "properties": {
    "data": {
      "title": "Data",
      "type": "string"
    },
    "graph_model_file": {
      "default": null,  <----
      "title": "Graph Model File",
      "type": "string"
    },
    "graph_model_name": {
      "default": null, <----
      "title": "Graph Model Name",
      "type": "string"
    }
  },
  "required": [
    "data"
  ],
  "title": "cognifyArguments"
}

But from the users perspective they don't look like null

If I call the tool with this values:

Error executing tool cognify: 2 validation errors for cognifyArguments
graph_model_file
  Input should be a valid string [type=string_type, input_value=None, input_type=NoneType]
    For further information visit https://errors.pydantic.dev/2.10/v/string_type
graph_model_name
  Input should be a valid string [type=string_type, input_value=None, input_type=NoneType]
    For further information visit https://errors.pydantic.dev/2.10/v/string_type

If I call the tool with raw mode and explicitly specify empty strings:

So generally speaking, this is a tool for developers, sometimes more control is a good thing! In this case it gives more information and let me pass exact value. In other case you may want to paste your parameters from IDE, I don't see it as redundant thing.

If you want something to be cahnged in my implementation, or add tests, I will be happy to do so.

[jjjuk]

Actually, if you type something in the textarea and then delete it, it will put "" into the parameters, but there is no obvious way to know it.
I figured it out when playing with raw mode)

Btw, Is this an expected behavior?