revert openai_compat changes and use OpenAIMixin for openai_chat_completion

tests/integration/inference/test_openai_completion.py

@@ -13,6 +13,13 @@ import pytest
 from ..test_cases.test_case import TestCase
 
 
+@pytest.fixture(autouse=True)
+def rate_limit_delay():
+    """Add delay between tests to avoid rate limiting from providers like Fireworks"""
+    yield
+    time.sleep(30)  # 30 second delay after each test
+
+
 def _normalize_text(text: str) -> str:
     """
     Normalize Unicode text by removing diacritical marks for comparison.

tests/integration/inference/test_openai_embeddings.py

@@ -6,6 +6,7 @@
 
 import base64
 import struct
+import time
 
 import pytest
 from openai import OpenAI
@@ -13,6 +14,13 @@ from openai import OpenAI
 from llama_stack.core.library_client import LlamaStackAsLibraryClient
 
 
+@pytest.fixture(autouse=True)
+def rate_limit_delay():
+    """Add delay between tests to avoid rate limiting from providers like Fireworks"""
+    yield
+    time.sleep(30)  # 30 second delay after each test
+
+
 def decode_base64_to_floats(base64_string: str) -> list[float]:
     """Helper function to decode base64 string to list of float32 values."""
     embedding_bytes = base64.b64decode(base64_string)

tests/integration/suites.py

@@ -112,9 +112,10 @@ SETUP_DEFINITIONS: dict[str, Setup] = {
         name="fireworks",
         description="Fireworks provider with a text model",
         defaults={
-            "text_model": "fireworks/accounts/fireworks/models/llama-v3p1-8b-instruct",
-            "vision_model": "fireworks/accounts/fireworks/models/llama-v3p2-90b-vision-instruct",
+            "text_model": "accounts/fireworks/models/llama-v3p1-8b-instruct",
+            "vision_model": "accounts/fireworks/models/llama-v3p2-90b-vision-instruct",
             "embedding_model": "nomic-ai/nomic-embed-text-v1.5",
+            # "embedding_model": "accounts/fireworks/models/qwen3-embedding-8b",
         },
     ),
 }