From 1616891c40c42fb87105b1d4a00c2539ba8fb9e0 Mon Sep 17 00:00:00 2001
From: Yiorgis Gozadinos <ggozadinos@gmail.com>
Date: Wed, 7 Jan 2026 18:46:50 +0200
Subject: [PATCH] Fix non-reasoning OpenAI models that do not support reasoning
 config

---
 CHANGELOG.md                      |  5 +++++
 haiku_rag_slim/haiku/rag/utils.py | 16 +++++++++++-----
 tests/test_utils.py               | 11 +++++++++++
 3 files changed, 27 insertions(+), 5 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index c6ee507d..7ebfbea7 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,6 +1,11 @@
 # Changelog
 ## [Unreleased]
 
+### Fixed
+
+- **OpenAI Non-Reasoning Models**: Fixed `reasoning_effort` parameter being sent to non-reasoning OpenAI models (gpt-4o, gpt-4o-mini), causing 400 errors. Now correctly detects reasoning models (o1, o3 series) using pydantic-ai's model profile.
+- **Bedrock Non-Reasoning Models**: Fixed same issue for OpenAI models on Bedrock.
+
 ## [0.24.0] - 2026-01-07
 
 ### Added
diff --git a/haiku_rag_slim/haiku/rag/utils.py b/haiku_rag_slim/haiku/rag/utils.py
index 060b3367..f6437e43 100644
--- a/haiku_rag_slim/haiku/rag/utils.py
+++ b/haiku_rag_slim/haiku/rag/utils.py
@@ -2,7 +2,7 @@ import sys
 from datetime import UTC, datetime
 from importlib import metadata
 from pathlib import Path
-from typing import TYPE_CHECKING, Any
+from typing import TYPE_CHECKING, Any, cast
 
 from dateutil import parser as dateutil_parser
 from packaging.version import Version, parse
@@ -145,10 +145,16 @@ def get_model(
         )
 
     elif provider == "openai":
+        from pydantic_ai.profiles.openai import OpenAIModelProfile, openai_model_profile
+
         openai_settings: Any = None
 
-        # Apply thinking control
-        if model_config.enable_thinking is not None:
+        # Apply thinking control only for reasoning models (o-series, gpt-5)
+        profile = cast(OpenAIModelProfile, openai_model_profile(model))
+        if (
+            model_config.enable_thinking is not None
+            and profile.openai_supports_encrypted_reasoning_content
+        ):
             if model_config.enable_thinking is False:
                 openai_settings = OpenAIChatModelSettings(openai_reasoning_effort="low")
             else:
@@ -247,8 +253,8 @@ def get_model(
                     }
                 else:
                     additional_fields = {"thinking": {"type": "disabled"}}
-            elif "gpt" in model or "o1" in model or "o3" in model:
-                # OpenAI models on Bedrock
+            elif "o1" in model or "o3" in model:
+                # OpenAI reasoning models on Bedrock (o-series only, not gpt-4o)
                 additional_fields = {
                     "reasoning_effort": "high"
                     if model_config.enable_thinking
diff --git a/tests/test_utils.py b/tests/test_utils.py
index 2005d18e..9afbe48c 100644
--- a/tests/test_utils.py
+++ b/tests/test_utils.py
@@ -176,6 +176,17 @@ def test_get_model_openai_with_thinking():
     assert isinstance(result, OpenAIChatModel)
 
 
+def test_get_model_openai_non_reasoning_model_ignores_thinking():
+    """Test that non-reasoning OpenAI models don't get reasoning_effort setting."""
+    model_config = ModelConfig(
+        provider="openai", name="gpt-4o-mini", enable_thinking=False
+    )
+    result = get_model(model_config)
+    assert isinstance(result, OpenAIChatModel)
+    # Non-reasoning models should not have reasoning_effort set
+    assert result._settings is None
+
+
 @pytest.mark.skipif(not HAS_ANTHROPIC, reason="Anthropic not installed")
 def test_get_model_anthropic():
     """Test get_model returns AnthropicModel for Anthropic."""