[GH-ISSUE #40] Add Pydantic model support to LLM schema handler #8

New issue

Closed

opened 2026-03-02 04:07:47 +03:00 by kerem · 1 comment

kerem commented

2026-03-02 04:07:47 +03:00

Owner

Originally created by @gadievron on GitHub (Dec 22, 2025).
Original GitHub issue: https://github.com/gadievron/raptor/issues/40

Problem

The function _dict_schema_to_pydantic() in packages/llm_analysis/llm/providers.py only handles dict schemas, but RAPTOR's hybrid design allows both dicts and Pydantic models. When Instructor library passes Pydantic models directly, TypeError occurs, breaking LLM structured output generation.

Root Cause

RAPTOR intentionally uses hybrid Pydantic/dict approach for flexibility, but current implementation doesn't detect when Pydantic model is passed.

Impact

TypeError when Instructor passes Pydantic models
Breaks LLM structured output generation
Limits flexibility of hybrid approach

Fix

Add Pydantic model detection:

from inspect import isclass

def _dict_schema_to_pydantic(schema: Union[Dict[str, Any], Type[BaseModel]]) -> Type[BaseModel]:
    if isclass(schema) and issubclass(schema, BaseModel):
        return schema
    if not isinstance(schema, dict):
        raise ValueError(f"Schema must be dict or Pydantic BaseModel class")
    # ... rest of dict processing ...

File: packages/llm_analysis/llm/providers.py:88-95

Also requires type annotation updates across codebase.

Type

Bug fix + Enhancement
Backward compatible

Bug #2 from internal analysis
Supports RAPTOR's "flexibility over strictness" philosophy

Originally created by @gadievron on GitHub (Dec 22, 2025). Original GitHub issue: https://github.com/gadievron/raptor/issues/40 ## Problem The function `_dict_schema_to_pydantic()` in `packages/llm_analysis/llm/providers.py` only handles dict schemas, but RAPTOR's hybrid design allows both dicts and Pydantic models. When Instructor library passes Pydantic models directly, TypeError occurs, breaking LLM structured output generation. ## Root Cause RAPTOR intentionally uses hybrid Pydantic/dict approach for flexibility, but current implementation doesn't detect when Pydantic model is passed. ## Impact - TypeError when Instructor passes Pydantic models - Breaks LLM structured output generation - Limits flexibility of hybrid approach ## Fix Add Pydantic model detection: ```python from inspect import isclass def _dict_schema_to_pydantic(schema: Union[Dict[str, Any], Type[BaseModel]]) -> Type[BaseModel]: if isclass(schema) and issubclass(schema, BaseModel): return schema if not isinstance(schema, dict): raise ValueError(f"Schema must be dict or Pydantic BaseModel class") # ... rest of dict processing ... ``` **File:** `packages/llm_analysis/llm/providers.py:88-95` Also requires type annotation updates across codebase. ## Type - Bug fix + Enhancement - Backward compatible ## Related - Bug #2 from internal analysis - Supports RAPTOR's "flexibility over strictness" philosophy