BaseTool Pydantic v1 Compatibility Fix

What it does

This patch restores full compatibility between LangChain’s BaseTool / Tool classes and Pydantic v1. Prior to this fix, creating a BaseTool subclass using Pydantic v1-style field declarations (Field(default=None) without an explicit annotation, or validator decorators) would raise a ValidationError or silently produce broken tool schemas at runtime. The fix patches the internal create_schema logic in langchain-core/tools/runnable.py to handle both Pydantic v1 and v2 field representations.

Why it matters

A large proportion of production deployments are still on Pydantic v1 because upgrading to v2 requires code changes or depends on other libraries. Those deployments hit cryptic validation errors when defining custom tools with structured inputs. After the fix, you can define tools with Field regardless of the installed Pydantic version.

Example

$ Inspect a tool’s auto-generated schema

=== Tool Schema ===
{
  "$defs": {
    "summarize_documentInput": {
      "properties": {
        "text": {"description": "The document text to summarize", "type": "string"},
        "max_length": {"default": 200, "description": "Max summary length in words", "type": "integer"}
      },
      "required": ["text"],
      "type": "object"
    }
  }
}

=== Tool Result ===
[Summary (20 words)]: LangChain is a framework for building applications...

Common flags

Field(default=200)

Pydantic v1-style default field declarations

args_schema

Explicit Pydantic model for tool input validation