Enable GPT-5

Author: NolanTrem

py/core/providers/llm/openai.py

@@ -399,14 +399,13 @@ def _get_base_args(self, generation_config: GenerationConfig) -> dict:
 
         model_str = generation_config.model or ""
 
-        if "o1" not in model_str and "o3" not in model_str:
+        if any(model_prefix in model_str.lower() for model_prefix in ["o1", "o3", "gpt-5"]):
+            args["max_completion_tokens"] = generation_config.max_tokens_to_sample
+
+        else:
             args["max_tokens"] = generation_config.max_tokens_to_sample
             args["temperature"] = generation_config.temperature
             args["top_p"] = generation_config.top_p
-        else:
-            args["max_completion_tokens"] = (
-                generation_config.max_tokens_to_sample
-            )
 
         if generation_config.reasoning_effort is not None:
             args["reasoning_effort"] = generation_config.reasoning_effort

py/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "r2r"
-version = "3.6.5"
+version = "3.6.6"
 description = "SciPhi R2R"
 readme = "README.md"
 license = {text = "MIT"}
@@ -17,7 +17,7 @@ dependencies = [
     "alembic >=1.13.3,<2.0.0",
     "fastapi >=0.115.11,<0.116.0",
     "httpx >=0.27.0",
-    "openai >=1.61.0",
+    "openai >=1.99.0",
     "python-dotenv >=1.0.1,<2.0.0",
     "psycopg-binary >=3.2.3,<4.0.0",
     "requests >=2.31.0,<3.0.0",

py/r2r/r2r.toml

@@ -9,13 +9,13 @@ default_max_collections_per_user = 5_000
 default_max_upload_size = 214748364800
 
 # LLM used for internal operations, like deriving conversation names
-fast_llm = "openai/gpt-4.1-mini"
+fast_llm = "openai/gpt-5-nano-2025-08-07"
 
 # LLM used for user-facing output, like RAG replies
-quality_llm = "openai/gpt-4.1"
+quality_llm = "openai/gpt-5-2025-08-07"
 
 # LLM used for ingesting visual inputs
-vlm = "openai/gpt-4.1"
+vlm = "openai/gpt-5-2025-08-07"
 
 # LLM used for transcription
 audio_lm = "openai/whisper-1"