Schema Generation Fails with 'type' Error When Using Nested Pydantic Models

[Fgulbahar]

Describe the bug

There's an error in the schema generation function (pydantic_model_to_json_schema) when handling Pydantic models. The error occurs in the clean_property function with the message: "Schema generation failed: Unexpected error in schema generation: 'type'". This happens when trying to create custom tool with nested models within the custom tool code sent - the system tries to generate JSON schemas for nested Pydantic models.

Please describe your setup

• How did you install letta?
  • docker
• Describe your setup
  • MacOS

Screenshots
Reproduction Steps
Here's a minimal example that reproduces the error:

from pydantic import BaseModel, Field

class DateFormatted(BaseModel):
    date: str = Field(..., description='The date string to be formatted')
    timezone: str = Field(..., description='The timezone to use for the date')
    date_format: str = Field(..., description='The format to apply to the date')

class WorkflowParams(BaseModel):
    date_formatted: DateFormatted = Field(
        ..., description='Format configuration for the date', title='DateFormatted'
    )
    valid: bool = Field(
        ..., description='If it is a valid value'
    )

When trying to generate the schema for these models, the error occurs in the clean_property function, suggesting that some property types aren't being handled correctly.

{ "detail": "Schema generation failed: Unexpected error in schema generation: 'type'" }

Additional context

The issue appears to be in the schema generation code where it doesn't properly handle all possible type structures that Pydantic can generate. The clean_property function needs to be more robust in handling:
Missing type fields
Complex type definitions
Nested model references

Letta Config
Using default configuration.