removed deprecated models

tee_gateway/controllers/chat_controller.py

@@ -39,6 +39,10 @@
     create_image_generation_response,
     create_image_generation_streaming_response,
 )
+from tee_gateway.model_errors import (
+    is_upstream_model_not_served,
+    upstream_model_not_served_payload,
+)
 from tee_gateway.model_registry import get_model_config
 from tee_gateway.pricing import compute_session_cost
 
@@ -367,6 +371,8 @@ def _create_non_streaming_response(chat_request: CreateChatCompletionRequest):
 
     except Exception as e:
         logger.error(f"Chat completion error: {str(e)}", exc_info=True)
+        if is_upstream_model_not_served(e):
+            return upstream_model_not_served_payload(chat_request.model, e), 404
         return {
             "error": str(e) or "Request processing failed",
             "exception_type": type(e).__name__,
@@ -805,10 +811,15 @@ def generate():
                 # event IS the terminal marker for the error path — we do NOT
                 # emit a trailing `[DONE]`, which conventionally signals a
                 # clean completion and would mis-signal an errored stream.
-                error_payload = {
-                    "error": str(e) or "Stream processing failed",
-                    "exception_type": type(e).__name__,
-                }
+                if is_upstream_model_not_served(e):
+                    error_payload = upstream_model_not_served_payload(
+                        chat_request.model, e
+                    )
+                else:
+                    error_payload = {
+                        "error": str(e) or "Stream processing failed",
+                        "exception_type": type(e).__name__,
+                    }
                 yield f"data: {json.dumps(error_payload)}\n\n"
 
         return Response(
@@ -822,6 +833,8 @@ def generate():
 
     except Exception as e:
         logger.error(f"Stream setup error: {str(e)}", exc_info=True)
+        if is_upstream_model_not_served(e):
+            return upstream_model_not_served_payload(chat_request.model, e), 404
         return {
             "error": str(e) or "Stream setup failed",
             "exception_type": type(e).__name__,

tee_gateway/controllers/completions_controller.py

@@ -18,6 +18,10 @@
     extract_web_search_count,
     extract_usage,
 )
+from tee_gateway.model_errors import (
+    is_upstream_model_not_served,
+    upstream_model_not_served_payload,
+)
 from tee_gateway.pricing import compute_session_cost
 
 logger = logging.getLogger(__name__)
@@ -116,6 +120,8 @@ def create_completion(body):
 
     except Exception as e:
         logger.error(f"Completion error: {str(e)}", exc_info=True)
+        if is_upstream_model_not_served(e):
+            return upstream_model_not_served_payload(body.model, e), 404
         return {
             "error": str(e) or "Request processing failed",
             "exception_type": type(e).__name__,

tee_gateway/image_generation.py

@@ -41,6 +41,10 @@
 from tee_gateway.models.create_chat_completion_response import (
     CreateChatCompletionResponse,
 )
+from tee_gateway.model_errors import (
+    is_upstream_model_not_served,
+    upstream_model_not_served_payload,
+)
 from tee_gateway.model_registry import get_model_config
 from tee_gateway.pricing import compute_session_cost
 from tee_gateway.tee_manager import get_tee_keys, compute_tee_msg_hash
@@ -498,10 +502,13 @@ def generate():
             # Surface the real exception detail to the client (matching the chat
             # streaming path) so browser-side logs show the actual cause instead
             # of an opaque generic string.
-            error_payload = {
-                "error": str(e) or "Stream processing failed",
-                "exception_type": type(e).__name__,
-            }
+            if is_upstream_model_not_served(e):
+                error_payload = upstream_model_not_served_payload(chat_request.model, e)
+            else:
+                error_payload = {
+                    "error": str(e) or "Stream processing failed",
+                    "exception_type": type(e).__name__,
+                }
             yield f"data: {json.dumps(error_payload)}\n\n"
 
     return Response(

tee_gateway/model_errors.py

@@ -0,0 +1,81 @@
+"""Helpers for turning provider model failures into client-facing errors."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from tee_gateway.model_registry import get_model_config
+
+
+def exception_status_code(exc: Exception) -> int | None:
+    """Extract an HTTP/gRPC-ish status code from common provider SDK exceptions."""
+    status = getattr(exc, "status_code", None)
+    if isinstance(status, int):
+        return status
+
+    response = getattr(exc, "response", None)
+    response_status = getattr(response, "status_code", None)
+    if isinstance(response_status, int):
+        return response_status
+
+    code_attr = getattr(exc, "code", None)
+    code_value: Any
+    if callable(code_attr):
+        try:
+            code_value = code_attr()
+        except Exception:
+            code_value = None
+    else:
+        code_value = code_attr
+
+    if isinstance(code_value, int):
+        return code_value
+    if getattr(code_value, "name", None) == "NOT_FOUND":
+        return 404
+
+    return None
+
+
+def is_upstream_model_not_served(exc: Exception) -> bool:
+    """Return True when a provider says a registered model is not available."""
+    if exception_status_code(exc) == 404:
+        return True
+
+    exc_type = type(exc).__name__.lower()
+    if "notfound" in exc_type or "not_found" in exc_type:
+        return True
+
+    message = str(exc).lower()
+    model_missing_phrases = (
+        "model not found",
+        "model is not found",
+        "model does not exist",
+        "model_not_found",
+        "not found for model",
+        "not served",
+        "is not supported for",
+    )
+    return any(phrase in message for phrase in model_missing_phrases)
+
+
+def upstream_model_not_served_payload(model: str, exc: Exception) -> dict[str, Any]:
+    """Build an OpenAI-style-ish payload for a provider model availability miss."""
+    payload: dict[str, Any] = {
+        "error": (
+            f"Upstream provider does not currently serve model '{model}'. "
+            "The model may have been removed, renamed, or disabled upstream."
+        ),
+        "code": "model_not_served_upstream",
+        "model": model,
+        "exception_type": type(exc).__name__,
+        "upstream_error": str(exc) or type(exc).__name__,
+    }
+
+    try:
+        cfg = get_model_config(model)
+    except Exception:
+        return payload
+
+    payload["provider"] = cfg.provider
+    payload["api_name"] = cfg.api_name
+    return payload

tee_gateway/model_registry.py

@@ -287,30 +287,6 @@ class SupportedModel(Enum):
     )
 
     # ── Google Gemini ───────────────────────────────────────────────────
-    # Note: gemini-2.5-flash, gemini-2.5-pro, and gemini-2.5-flash-lite are scheduled
-    # for deprecation on June 17, 2026 (flash-lite: July 22, 2026). Use the Gemini 3
-    # replacements below for new integrations.
-    GEMINI_2_5_FLASH = ModelConfig(
-        provider="google",
-        api_name="gemini-2.5-flash",
-        input_price_usd=Decimal("0.0000003"),
-        output_price_usd=Decimal("0.0000025"),
-        thinking_budget=0,
-    )
-    GEMINI_2_5_PRO = ModelConfig(
-        provider="google",
-        api_name="gemini-2.5-pro",
-        input_price_usd=Decimal("0.00000125"),
-        output_price_usd=Decimal("0.00001"),
-        thinking_budget=128,
-    )
-    GEMINI_2_5_FLASH_LITE = ModelConfig(
-        provider="google",
-        api_name="gemini-2.5-flash-lite",
-        input_price_usd=Decimal("0.0000001"),
-        output_price_usd=Decimal("0.0000004"),
-        thinking_budget=0,
-    )
     GEMINI_3_FLASH_PREVIEW = ModelConfig(
         provider="google",
         api_name="gemini-3-flash-preview",
@@ -324,9 +300,9 @@ class SupportedModel(Enum):
         output_price_usd=Decimal("0.000012"),
         thinking_budget=128,
     )
-    GEMINI_3_1_FLASH_LITE_PREVIEW = ModelConfig(
+    GEMINI_3_1_FLASH_LITE = ModelConfig(
         provider="google",
-        api_name="gemini-3.1-flash-lite-preview",
+        api_name="gemini-3.1-flash-lite",
         input_price_usd=Decimal("0.00000025"),
         output_price_usd=Decimal("0.0000015"),
         thinking_budget=0,
@@ -374,53 +350,23 @@ class SupportedModel(Enum):
         input_price_usd=Decimal("0.00000125"),
         output_price_usd=Decimal("0.0000025"),
     )
-    GROK_4 = ModelConfig(
-        provider="x-ai",
-        api_name="grok-4",
-        input_price_usd=Decimal("0.000003"),
-        output_price_usd=Decimal("0.000015"),
-    )
-    GROK_4_FAST = ModelConfig(
-        provider="x-ai",
-        api_name="grok-4-fast",
-        input_price_usd=Decimal("0.0000002"),
-        output_price_usd=Decimal("0.0000005"),
-    )
-    GROK_4_1_FAST = ModelConfig(
-        provider="x-ai",
-        api_name="grok-4-1-fast",
-        input_price_usd=Decimal("0.0000002"),
-        output_price_usd=Decimal("0.0000005"),
-    )
-    GROK_4_1_FAST_NON_REASONING = ModelConfig(
-        provider="x-ai",
-        api_name="grok-4-1-fast-non-reasoning",
-        input_price_usd=Decimal("0.0000002"),
-        output_price_usd=Decimal("0.0000005"),
-    )
     GROK_4_20_REASONING = ModelConfig(
         provider="x-ai",
-        api_name="grok-4.20-reasoning",
+        api_name="grok-4.20-0309-reasoning",
         input_price_usd=Decimal("0.000002"),
         output_price_usd=Decimal("0.000006"),
     )
     GROK_4_20_NON_REASONING = ModelConfig(
         provider="x-ai",
-        api_name="grok-4.20-non-reasoning",
+        api_name="grok-4.20-0309-non-reasoning",
         input_price_usd=Decimal("0.000002"),
         output_price_usd=Decimal("0.000006"),
     )
-    GROK_CODE_FAST_1 = ModelConfig(
-        provider="x-ai",
-        api_name="grok-code-fast-1",
-        input_price_usd=Decimal("0.0000002"),
-        output_price_usd=Decimal("0.0000015"),
-    )
     # Image generation via xAI's OpenAI-compatible /images/generations endpoint
     # (Aurora). Billed at a flat $0.07 per generated image; token prices unused.
     GROK_2_IMAGE = ModelConfig(
         provider="x-ai",
-        api_name="grok-2-image-1212",
+        api_name="grok-imagine-image",
         input_price_usd=Decimal("0"),
         output_price_usd=Decimal("0"),
         image_generation=True,
@@ -554,20 +500,6 @@ class SupportedModel(Enum):
         image_extra_params={"size": "1280x1280"},
     )
 
-    # ── Legacy models (not in current SDK — retained for older SDK versions) ──
-    GROK_3_MINI = ModelConfig(
-        provider="x-ai",
-        api_name="grok-3-mini",
-        input_price_usd=Decimal("0.0000003"),
-        output_price_usd=Decimal("0.0000005"),
-    )
-    GROK_3 = ModelConfig(
-        provider="x-ai",
-        api_name="grok-3-latest",
-        input_price_usd=Decimal("0.000003"),
-        output_price_usd=Decimal("0.000015"),
-    )
-
 
 # Canonical lookup: user-facing model name → SupportedModel
 # The "user-facing name" is what callers pass in the `model` field of requests.
@@ -605,30 +537,20 @@ class SupportedModel(Enum):
     "claude-opus-4-8": SupportedModel.CLAUDE_OPUS_4_8,
     "claude-fable-5": SupportedModel.CLAUDE_FABLE_5,
     # Google
-    "gemini-2.5-flash": SupportedModel.GEMINI_2_5_FLASH,
-    "gemini-2.5-pro": SupportedModel.GEMINI_2_5_PRO,
-    "gemini-2.5-flash-lite": SupportedModel.GEMINI_2_5_FLASH_LITE,
     "gemini-3-flash-preview": SupportedModel.GEMINI_3_FLASH_PREVIEW,
     "gemini-3.1-pro-preview": SupportedModel.GEMINI_3_1_PRO_PREVIEW,
-    "gemini-3.1-flash-lite-preview": SupportedModel.GEMINI_3_1_FLASH_LITE_PREVIEW,
+    "gemini-3.1-flash-lite": SupportedModel.GEMINI_3_1_FLASH_LITE,
     "gemini-2.5-flash-image": SupportedModel.GEMINI_2_5_FLASH_IMAGE,
     "gemini-3.1-flash-image": SupportedModel.GEMINI_3_1_FLASH_IMAGE,
     "gemini-3.5-flash": SupportedModel.GEMINI_3_5_FLASH,
     # xAI
     "grok-4.5": SupportedModel.GROK_4_5,
     "grok-4.5-latest": SupportedModel.GROK_4_5,
     "grok-4.3": SupportedModel.GROK_4_3,
-    "grok-4": SupportedModel.GROK_4,
-    "grok-4-fast": SupportedModel.GROK_4_FAST,
-    "grok-4-1-fast": SupportedModel.GROK_4_1_FAST,
-    "grok-4.1-fast": SupportedModel.GROK_4_1_FAST,
-    "grok-4-1-fast-non-reasoning": SupportedModel.GROK_4_1_FAST_NON_REASONING,
     "grok-4.20-reasoning": SupportedModel.GROK_4_20_REASONING,
     "grok-4.20-non-reasoning": SupportedModel.GROK_4_20_NON_REASONING,
-    "grok-code-fast-1": SupportedModel.GROK_CODE_FAST_1,
+    "grok-imagine-image": SupportedModel.GROK_2_IMAGE,
     "grok-2-image": SupportedModel.GROK_2_IMAGE,
-    "grok-2-image-1212": SupportedModel.GROK_2_IMAGE,
-    "grok-2-image-latest": SupportedModel.GROK_2_IMAGE,
     # ByteDance
     "seed-1-6-250615": SupportedModel.SEED_1_6,
     "seed-1.6": SupportedModel.SEED_1_6,
@@ -656,11 +578,6 @@ class SupportedModel(Enum):
     # Z.ai
     "glm-5.2": SupportedModel.GLM_5_2,
     "glm-image": SupportedModel.GLM_IMAGE,
-    # Legacy — not in current SDK, retained for older SDK versions
-    "grok-3-mini-beta": SupportedModel.GROK_3_MINI,  # old beta alias
-    "grok-3-mini": SupportedModel.GROK_3_MINI,
-    "grok-3-beta": SupportedModel.GROK_3,  # old beta alias
-    "grok-3": SupportedModel.GROK_3,
 }
 
 # Build the rate card automatically from the enum (for backward compat with util.py)

tee_gateway/test/test_tee_core.py

@@ -445,8 +445,8 @@ def test_all_four_providers_reachable(self):
         providers = {
             get_model_config("gpt-4.1").provider,
             get_model_config("claude-sonnet-4-5").provider,
-            get_model_config("gemini-2.5-flash").provider,
-            get_model_config("grok-4").provider,
+            get_model_config("gemini-3.5-flash").provider,
+            get_model_config("grok-4.3").provider,
         }
         self.assertEqual(providers, {"openai", "anthropic", "google", "x-ai"})
 
@@ -460,11 +460,11 @@ def test_anthropic_model_lookup(self):
         self.assertEqual(cfg.provider, "anthropic")
 
     def test_google_model_lookup(self):
-        cfg = get_model_config("gemini-2.5-flash")
+        cfg = get_model_config("gemini-3.5-flash")
         self.assertEqual(cfg.provider, "google")
 
     def test_xai_model_lookup(self):
-        cfg = get_model_config("grok-4")
+        cfg = get_model_config("grok-4.3")
         self.assertEqual(cfg.provider, "x-ai")
 
     def test_unknown_model_raises_value_error(self):
@@ -493,7 +493,7 @@ def test_model_alias_and_dated_name_resolve_identically(self):
     def test_models_with_force_temperature_have_positive_value(self):
         """Any model that forces a temperature must set it to a positive float.
         We don't pin the exact value — just confirm the field is plausible if set."""
-        cfg = get_model_config("o4-mini")
+        cfg = get_model_config("o3")
         if cfg.force_temperature is not None:
             self.assertGreater(cfg.force_temperature, 0)
 
@@ -810,7 +810,7 @@ def test_plain_text_request_passes(self):
     def test_image_blocked_when_model_lacks_support(self):
         with mock.patch(self.CAPS, return_value={"image_inputs": False}):
             with self.assertRaises(AttachmentValidationError):
-                validate_attachments(self._image_msg("aGVsbG8="), "grok-4")
+                validate_attachments(self._image_msg("aGVsbG8="), "grok-4.3")
 
     def test_image_allowed_when_model_supports(self):
         with mock.patch(self.CAPS, return_value={"image_inputs": True}):
@@ -826,7 +826,7 @@ def test_pdf_blocked_when_model_lacks_support(self):
             self.CAPS, return_value={"image_inputs": True, "pdf_inputs": False}
         ):
             with self.assertRaises(AttachmentValidationError):
-                validate_attachments(self._pdf_msg("JVBERi0="), "grok-4")
+                validate_attachments(self._pdf_msg("JVBERi0="), "grok-4.3")
 
 
 # ---------------------------------------------------------------------------