refactor: moved registries into a separate module and code cleanup

fix: refactored dial provider to follow the same pattern
2025-10-07 12:59:09 +04:00
parent c27e81d6d2
commit 7c36b9255a
54 changed files with 325 additions and 282 deletions
--- a/conf/dial_models.json
+++ b/conf/dial_models.json
@@ -0,0 +1,169 @@
+{
+  "_README": {
+    "description": "Model metadata for the DIAL (Data & AI Layer) aggregation provider.",
+    "documentation": "https://github.com/BeehiveInnovations/zen-mcp-server/blob/main/docs/configuration.md",
+    "usage": "Models listed here are exposed through the DIAL provider. Aliases are case-insensitive.",
+    "field_notes": "Matches providers/shared/model_capabilities.py.",
+    "field_descriptions": {
+      "model_name": "The model identifier as exposed by DIAL (typically deployment name)",
+      "aliases": "Array of shorthand names users can type instead of the full model name",
+      "context_window": "Total number of tokens the model can process (input + output combined)",
+      "max_output_tokens": "Maximum number of tokens the model can generate in a single response",
+      "supports_extended_thinking": "Whether the model supports extended reasoning tokens",
+      "supports_json_mode": "Whether the model can guarantee valid JSON output",
+      "supports_function_calling": "Whether the model supports function/tool calling",
+      "supports_images": "Whether the model can process images/visual input",
+      "max_image_size_mb": "Maximum total size in MB for all images combined",
+      "supports_temperature": "Whether the model accepts the temperature parameter",
+      "temperature_constraint": "Temperature constraint hint: 'fixed', 'range', or 'discrete'",
+      "description": "Human-readable description of the model",
+      "intelligence_score": "1-20 human rating used as the primary signal for auto-mode ordering"
+    }
+  },
+  "models": [
+    {
+      "model_name": "o3-2025-04-16",
+      "friendly_name": "DIAL (O3)",
+      "aliases": ["o3"],
+      "intelligence_score": 14,
+      "description": "OpenAI O3 via DIAL - Strong reasoning model",
+      "context_window": 200000,
+      "max_output_tokens": 100000,
+      "supports_extended_thinking": false,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": false,
+      "temperature_constraint": "fixed"
+    },
+    {
+      "model_name": "o4-mini-2025-04-16",
+      "friendly_name": "DIAL (O4-mini)",
+      "aliases": ["o4-mini"],
+      "intelligence_score": 11,
+      "description": "OpenAI O4-mini via DIAL - Fast reasoning model",
+      "context_window": 200000,
+      "max_output_tokens": 100000,
+      "supports_extended_thinking": false,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": false,
+      "temperature_constraint": "fixed"
+    },
+    {
+      "model_name": "anthropic.claude-sonnet-4.1-20250805-v1:0",
+      "friendly_name": "DIAL (Sonnet 4.1)",
+      "aliases": ["sonnet-4.1", "sonnet-4"],
+      "intelligence_score": 10,
+      "description": "Claude Sonnet 4.1 via DIAL - Balanced performance",
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 5.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range"
+    },
+    {
+      "model_name": "anthropic.claude-sonnet-4.1-20250805-v1:0-with-thinking",
+      "friendly_name": "DIAL (Sonnet 4.1 Thinking)",
+      "aliases": ["sonnet-4.1-thinking", "sonnet-4-thinking"],
+      "intelligence_score": 11,
+      "description": "Claude Sonnet 4.1 with thinking mode via DIAL",
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": true,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 5.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range"
+    },
+    {
+      "model_name": "anthropic.claude-opus-4.1-20250805-v1:0",
+      "friendly_name": "DIAL (Opus 4.1)",
+      "aliases": ["opus-4.1", "opus-4"],
+      "intelligence_score": 14,
+      "description": "Claude Opus 4.1 via DIAL - Most capable Claude model",
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": false,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 5.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range"
+    },
+    {
+      "model_name": "anthropic.claude-opus-4.1-20250805-v1:0-with-thinking",
+      "friendly_name": "DIAL (Opus 4.1 Thinking)",
+      "aliases": ["opus-4.1-thinking", "opus-4-thinking"],
+      "intelligence_score": 15,
+      "description": "Claude Opus 4.1 with thinking mode via DIAL",
+      "context_window": 200000,
+      "max_output_tokens": 64000,
+      "supports_extended_thinking": true,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 5.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range"
+    },
+    {
+      "model_name": "gemini-2.5-pro-preview-03-25-google-search",
+      "friendly_name": "DIAL (Gemini 2.5 Pro Search)",
+      "aliases": ["gemini-2.5-pro-search"],
+      "intelligence_score": 17,
+      "description": "Gemini 2.5 Pro with Google Search via DIAL",
+      "context_window": 1000000,
+      "max_output_tokens": 65536,
+      "supports_extended_thinking": false,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range"
+    },
+    {
+      "model_name": "gemini-2.5-pro-preview-05-06",
+      "friendly_name": "DIAL (Gemini 2.5 Pro)",
+      "aliases": ["gemini-2.5-pro"],
+      "intelligence_score": 18,
+      "description": "Gemini 2.5 Pro via DIAL - Deep reasoning",
+      "context_window": 1000000,
+      "max_output_tokens": 65536,
+      "supports_extended_thinking": false,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range"
+    },
+    {
+      "model_name": "gemini-2.5-flash-preview-05-20",
+      "friendly_name": "DIAL (Gemini Flash 2.5)",
+      "aliases": ["gemini-2.5-flash"],
+      "intelligence_score": 10,
+      "description": "Gemini 2.5 Flash via DIAL - Ultra-fast",
+      "context_window": 1000000,
+      "max_output_tokens": 65536,
+      "supports_extended_thinking": false,
+      "supports_function_calling": false,
+      "supports_json_mode": true,
+      "supports_images": true,
+      "max_image_size_mb": 20.0,
+      "supports_temperature": true,
+      "temperature_constraint": "range"
+    }
+  ]
+}
--- a/conf/openai_models.json
+++ b/conf/openai_models.json
@@ -53,7 +53,7 @@
        "gpt5-pro"
      ],
      "intelligence_score": 18,
-      "description": "GPT-5 Pro (400K context, 272K output) - Advanced model with reasoning support",
+      "description": "GPT-5 Pro (400K context, 272K output) - Very advanced, reasoning model",
      "context_window": 400000,
      "max_output_tokens": 272000,
      "supports_extended_thinking": true,
@@ -156,7 +156,7 @@
        "o3pro"
      ],
      "intelligence_score": 15,
-      "description": "Professional-grade reasoning (200K context) - EXTREMELY EXPENSIVE: Only for the most complex problems requiring universe-scale complexity analysis OR when the user explicitly asks for this model. Use sparingly for critical architectural decisions or exceptionally complex debugging that other models cannot handle.",
+      "description": "Professional-grade reasoning (200K context)",
      "context_window": 200000,
      "max_output_tokens": 65536,
      "supports_extended_thinking": false,
--- a/config.py
+++ b/config.py
@@ -30,7 +30,8 @@ DEFAULT_MODEL = get_env("DEFAULT_MODEL", "auto") or "auto"
 # Auto mode detection - when DEFAULT_MODEL is "auto", Claude picks the model
 IS_AUTO_MODE = DEFAULT_MODEL.lower() == "auto"

-# Each provider (gemini.py, openai_provider.py, xai.py) defines its own MODEL_CAPABILITIES
+# Each provider (gemini.py, openai.py, xai.py, dial.py, openrouter.py, custom.py, azure_openai.py)
+# defines its own MODEL_CAPABILITIES
 # with detailed descriptions. Tools use ModelProviderRegistry.get_available_model_names()
 # to get models only from enabled providers (those with valid API keys).
 #
--- a/docs/configuration.md
+++ b/docs/configuration.md
@@ -72,6 +72,7 @@ DEFAULT_MODEL=auto  # Claude picks best model for each task (recommended)
  - `conf/gemini_models.json` – Gemini catalogue (`GEMINI_MODELS_CONFIG_PATH`)
  - `conf/xai_models.json` – X.AI / GROK catalogue (`XAI_MODELS_CONFIG_PATH`)
  - `conf/openrouter_models.json` – OpenRouter catalogue (`OPENROUTER_MODELS_CONFIG_PATH`)
+  - `conf/dial_models.json` – DIAL aggregation catalogue (`DIAL_MODELS_CONFIG_PATH`)
  - `conf/custom_models.json` – Custom/OpenAI-compatible endpoints (`CUSTOM_MODELS_CONFIG_PATH`)

  Each JSON file documents the allowed fields via its `_README` block and controls model aliases, capability limits, and feature flags. Edit these files (or point the matching `*_MODELS_CONFIG_PATH` variable to your own copy) when you want to adjust context windows, enable JSON mode, or expose additional aliases without touching Python code.
@@ -154,6 +155,7 @@ OPENAI_MODELS_CONFIG_PATH=/path/to/openai_models.json
 GEMINI_MODELS_CONFIG_PATH=/path/to/gemini_models.json
 XAI_MODELS_CONFIG_PATH=/path/to/xai_models.json
 OPENROUTER_MODELS_CONFIG_PATH=/path/to/openrouter_models.json
+DIAL_MODELS_CONFIG_PATH=/path/to/dial_models.json
 CUSTOM_MODELS_CONFIG_PATH=/path/to/custom_models.json
 ```

--- a/docs/custom_models.md
+++ b/docs/custom_models.md
@@ -41,6 +41,7 @@ Zen ships multiple registries:
 - `conf/gemini_models.json` – native Google Gemini catalogue (`GEMINI_MODELS_CONFIG_PATH`)
 - `conf/xai_models.json` – native X.AI / GROK catalogue (`XAI_MODELS_CONFIG_PATH`)
 - `conf/openrouter_models.json` – OpenRouter catalogue (`OPENROUTER_MODELS_CONFIG_PATH`)
+- `conf/dial_models.json` – DIAL aggregation catalogue (`DIAL_MODELS_CONFIG_PATH`)
 - `conf/custom_models.json` – local/self-hosted OpenAI-compatible catalogue (`CUSTOM_MODELS_CONFIG_PATH`)

 Copy whichever file you need into your project (or point the corresponding `*_MODELS_CONFIG_PATH` env var at your own copy) and edit it to advertise the models you want.
@@ -71,7 +72,7 @@ Consult the JSON file for the full list, aliases, and capability flags. Add new

 View the baseline OpenRouter catalogue in [`conf/openrouter_models.json`](conf/openrouter_models.json) and populate [`conf/custom_models.json`](conf/custom_models.json) with your local models.

-Native catalogues (`conf/openai_models.json`, `conf/gemini_models.json`, `conf/xai_models.json`) follow the same schema. Updating those files lets you:
+Native catalogues (`conf/openai_models.json`, `conf/gemini_models.json`, `conf/xai_models.json`, `conf/dial_models.json`) follow the same schema. Updating those files lets you:

 - Expose new aliases (e.g., map `enterprise-pro` to `gpt-5-pro`)
 - Advertise support for JSON mode or vision if the upstream provider adds it
--- a/providers/init.py
+++ b/providers/init.py
@@ -3,8 +3,8 @@
 from .azure_openai import AzureOpenAIProvider
 from .base import ModelProvider
 from .gemini import GeminiModelProvider
+from .openai import OpenAIModelProvider
 from .openai_compatible import OpenAICompatibleProvider
-from .openai_provider import OpenAIModelProvider
 from .openrouter import OpenRouterProvider
 from .registry import ModelProviderRegistry
 from .shared import ModelCapabilities, ModelResponse
--- a/providers/azure_openai.py
+++ b/providers/azure_openai.py
@@ -12,9 +12,9 @@ except ImportError:  # pragma: no cover

 from utils.env import get_env, suppress_env_vars

-from .azure_registry import AzureModelRegistry
+from .openai import OpenAIModelProvider
 from .openai_compatible import OpenAICompatibleProvider
-from .openai_provider import OpenAIModelProvider
+from .registries.azure import AzureModelRegistry
 from .shared import ModelCapabilities, ModelResponse, ProviderType, TemperatureConstraint

 logger = logging.getLogger(__name__)
--- a/providers/custom.py
+++ b/providers/custom.py
@@ -4,11 +4,12 @@ import logging

 from utils.env import get_env

-from .custom_registry import CustomEndpointModelRegistry
 from .openai_compatible import OpenAICompatibleProvider
-from .openrouter_registry import OpenRouterModelRegistry
+from .registries.custom import CustomEndpointModelRegistry
+from .registries.openrouter import OpenRouterModelRegistry
 from .shared import ModelCapabilities, ProviderType

+
 class CustomProvider(OpenAICompatibleProvider):
    """Adapter for self-hosted or local OpenAI-compatible endpoints.

--- a/providers/dial.py
+++ b/providers/dial.py
@@ -2,17 +2,19 @@

 import logging
 import threading
-from typing import Optional
+from typing import ClassVar, Optional

 from utils.env import get_env

 from .openai_compatible import OpenAICompatibleProvider
-from .shared import ModelCapabilities, ModelResponse, ProviderType, TemperatureConstraint
+from .registries.dial import DialModelRegistry
+from .registry_provider_mixin import RegistryBackedProviderMixin
+from .shared import ModelCapabilities, ModelResponse, ProviderType

 logger = logging.getLogger(__name__)


-class DIALModelProvider(OpenAICompatibleProvider):
+class DIALModelProvider(RegistryBackedProviderMixin, OpenAICompatibleProvider):
    """Client for the DIAL (Data & AI Layer) aggregation service.

    DIAL exposes several third-party models behind a single OpenAI-compatible
@@ -23,185 +25,13 @@ class DIALModelProvider(OpenAICompatibleProvider):

    FRIENDLY_NAME = "DIAL"

+    REGISTRY_CLASS = DialModelRegistry
+    MODEL_CAPABILITIES: ClassVar[dict[str, ModelCapabilities]] = {}
+
    # Retry configuration for API calls
    MAX_RETRIES = 4
    RETRY_DELAYS = [1, 3, 5, 8]  # seconds

-    # Model configurations using ModelCapabilities objects
-    MODEL_CAPABILITIES = {
-        "o3-2025-04-16": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="o3-2025-04-16",
-            friendly_name="DIAL (O3)",
-            intelligence_score=14,
-            context_window=200_000,
-            max_output_tokens=100_000,
-            supports_extended_thinking=False,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,  # DIAL may not expose function calling
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=20.0,
-            supports_temperature=False,  # O3 models don't accept temperature
-            temperature_constraint=TemperatureConstraint.create("fixed"),
-            description="OpenAI O3 via DIAL - Strong reasoning model",
-            aliases=["o3"],
-        ),
-        "o4-mini-2025-04-16": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="o4-mini-2025-04-16",
-            friendly_name="DIAL (O4-mini)",
-            intelligence_score=11,
-            context_window=200_000,
-            max_output_tokens=100_000,
-            supports_extended_thinking=False,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,  # DIAL may not expose function calling
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=20.0,
-            supports_temperature=False,  # O4 models don't accept temperature
-            temperature_constraint=TemperatureConstraint.create("fixed"),
-            description="OpenAI O4-mini via DIAL - Fast reasoning model",
-            aliases=["o4-mini"],
-        ),
-        "anthropic.claude-sonnet-4.1-20250805-v1:0": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="anthropic.claude-sonnet-4.1-20250805-v1:0",
-            friendly_name="DIAL (Sonnet 4.1)",
-            intelligence_score=10,
-            context_window=200_000,
-            max_output_tokens=64_000,
-            supports_extended_thinking=False,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=5.0,
-            supports_temperature=True,
-            temperature_constraint=TemperatureConstraint.create("range"),
-            description="Claude Sonnet 4.1 via DIAL - Balanced performance",
-            aliases=["sonnet-4.1", "sonnet-4"],
-        ),
-        "anthropic.claude-sonnet-4.1-20250805-v1:0-with-thinking": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="anthropic.claude-sonnet-4.1-20250805-v1:0-with-thinking",
-            friendly_name="DIAL (Sonnet 4.1 Thinking)",
-            intelligence_score=11,
-            context_window=200_000,
-            max_output_tokens=64_000,
-            supports_extended_thinking=True,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=5.0,
-            supports_temperature=True,
-            temperature_constraint=TemperatureConstraint.create("range"),
-            description="Claude Sonnet 4.1 with thinking mode via DIAL",
-            aliases=["sonnet-4.1-thinking", "sonnet-4-thinking"],
-        ),
-        "anthropic.claude-opus-4.1-20250805-v1:0": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="anthropic.claude-opus-4.1-20250805-v1:0",
-            friendly_name="DIAL (Opus 4.1)",
-            intelligence_score=14,
-            context_window=200_000,
-            max_output_tokens=64_000,
-            supports_extended_thinking=False,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=5.0,
-            supports_temperature=True,
-            temperature_constraint=TemperatureConstraint.create("range"),
-            description="Claude Opus 4.1 via DIAL - Most capable Claude model",
-            aliases=["opus-4.1", "opus-4"],
-        ),
-        "anthropic.claude-opus-4.1-20250805-v1:0-with-thinking": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="anthropic.claude-opus-4.1-20250805-v1:0-with-thinking",
-            friendly_name="DIAL (Opus 4.1 Thinking)",
-            intelligence_score=15,
-            context_window=200_000,
-            max_output_tokens=64_000,
-            supports_extended_thinking=True,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=5.0,
-            supports_temperature=True,
-            temperature_constraint=TemperatureConstraint.create("range"),
-            description="Claude Opus 4.1 with thinking mode via DIAL",
-            aliases=["opus-4.1-thinking", "opus-4-thinking"],
-        ),
-        "gemini-2.5-pro-preview-03-25-google-search": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="gemini-2.5-pro-preview-03-25-google-search",
-            friendly_name="DIAL (Gemini 2.5 Pro Search)",
-            intelligence_score=17,
-            context_window=1_000_000,
-            max_output_tokens=65_536,
-            supports_extended_thinking=False,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=20.0,
-            supports_temperature=True,
-            temperature_constraint=TemperatureConstraint.create("range"),
-            description="Gemini 2.5 Pro with Google Search via DIAL",
-            aliases=["gemini-2.5-pro-search"],
-        ),
-        "gemini-2.5-pro-preview-05-06": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="gemini-2.5-pro-preview-05-06",
-            friendly_name="DIAL (Gemini 2.5 Pro)",
-            intelligence_score=18,
-            context_window=1_000_000,
-            max_output_tokens=65_536,
-            supports_extended_thinking=False,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=20.0,
-            supports_temperature=True,
-            temperature_constraint=TemperatureConstraint.create("range"),
-            description="Gemini 2.5 Pro via DIAL - Deep reasoning",
-            aliases=["gemini-2.5-pro"],
-        ),
-        "gemini-2.5-flash-preview-05-20": ModelCapabilities(
-            provider=ProviderType.DIAL,
-            model_name="gemini-2.5-flash-preview-05-20",
-            friendly_name="DIAL (Gemini Flash 2.5)",
-            intelligence_score=10,
-            context_window=1_000_000,
-            max_output_tokens=65_536,
-            supports_extended_thinking=False,
-            supports_system_prompts=True,
-            supports_streaming=True,
-            supports_function_calling=False,
-            supports_json_mode=True,
-            supports_images=True,
-            max_image_size_mb=20.0,
-            supports_temperature=True,
-            temperature_constraint=TemperatureConstraint.create("range"),
-            description="Gemini 2.5 Flash via DIAL - Ultra-fast",
-            aliases=["gemini-2.5-flash"],
-        ),
-    }
-
    def __init__(self, api_key: str, **kwargs):
        """Initialize DIAL provider with API key and host.

@@ -209,6 +39,7 @@ class DIALModelProvider(OpenAICompatibleProvider):
            api_key: DIAL API key for authentication
            **kwargs: Additional configuration options
        """
+        self._ensure_registry()
        # Get DIAL API host from environment or kwargs
        dial_host = kwargs.get("base_url") or get_env("DIAL_API_HOST") or "https://core.dialx.ai"

--- a/providers/gemini.py
+++ b/providers/gemini.py
@@ -2,7 +2,7 @@

 import base64
 import logging
-from typing import TYPE_CHECKING, Optional
+from typing import TYPE_CHECKING, ClassVar, Optional

 if TYPE_CHECKING:
    from tools.models import ToolModelCategory
@@ -14,7 +14,7 @@ from utils.env import get_env
 from utils.image_utils import validate_image

 from .base import ModelProvider
-from .gemini_registry import GeminiModelRegistry
+from .registries.gemini import GeminiModelRegistry
 from .registry_provider_mixin import RegistryBackedProviderMixin
 from .shared import ModelCapabilities, ModelResponse, ProviderType

@@ -30,7 +30,7 @@ class GeminiModelProvider(RegistryBackedProviderMixin, ModelProvider):
    """

    REGISTRY_CLASS = GeminiModelRegistry
-    MODEL_CAPABILITIES: dict[str, ModelCapabilities] = {}
+    MODEL_CAPABILITIES: ClassVar[dict[str, ModelCapabilities]] = {}

    # Thinking mode configurations - percentages of model's max_thinking_tokens
    # These percentages work across all models that support thinking
--- a/providers/openai_provider.py
+++ b/providers/openai_provider.py
@@ -1,13 +1,13 @@
 """OpenAI model provider implementation."""

 import logging
-from typing import TYPE_CHECKING, Optional
+from typing import TYPE_CHECKING, ClassVar, Optional

 if TYPE_CHECKING:
    from tools.models import ToolModelCategory

 from .openai_compatible import OpenAICompatibleProvider
-from .openai_registry import OpenAIModelRegistry
+from .registries.openai import OpenAIModelRegistry
 from .registry_provider_mixin import RegistryBackedProviderMixin
 from .shared import ModelCapabilities, ProviderType

@@ -23,7 +23,7 @@ class OpenAIModelProvider(RegistryBackedProviderMixin, OpenAICompatibleProvider)
    """

    REGISTRY_CLASS = OpenAIModelRegistry
-    MODEL_CAPABILITIES: dict[str, ModelCapabilities] = {}
+    MODEL_CAPABILITIES: ClassVar[dict[str, ModelCapabilities]] = {}

    def __init__(self, api_key: str, **kwargs):
        """Initialize OpenAI provider with API key."""
@@ -50,7 +50,7 @@ class OpenAIModelProvider(RegistryBackedProviderMixin, OpenAICompatibleProvider)
            return builtin

        try:
-            from .openrouter_registry import OpenRouterModelRegistry
+            from .registries.openrouter import OpenRouterModelRegistry

            registry = OpenRouterModelRegistry()
            config = registry.get_model_config(canonical_name)
--- a/providers/openrouter.py
+++ b/providers/openrouter.py
@@ -5,7 +5,7 @@ import logging
 from utils.env import get_env

 from .openai_compatible import OpenAICompatibleProvider
-from .openrouter_registry import OpenRouterModelRegistry
+from .registries.openrouter import OpenRouterModelRegistry
 from .shared import (
    ModelCapabilities,
    ProviderType,
--- a/providers/registries/init.py
+++ b/providers/registries/init.py
@@ -0,0 +1,19 @@
+"""Registry implementations for provider capability manifests."""
+
+from .azure import AzureModelRegistry
+from .custom import CustomEndpointModelRegistry
+from .dial import DialModelRegistry
+from .gemini import GeminiModelRegistry
+from .openai import OpenAIModelRegistry
+from .openrouter import OpenRouterModelRegistry
+from .xai import XAIModelRegistry
+
+__all__ = [
+    "AzureModelRegistry",
+    "CustomEndpointModelRegistry",
+    "DialModelRegistry",
+    "GeminiModelRegistry",
+    "OpenAIModelRegistry",
+    "OpenRouterModelRegistry",
+    "XAIModelRegistry",
+]
--- a/providers/registries/azure.py
+++ b/providers/registries/azure.py
@@ -4,8 +4,8 @@ from __future__ import annotations

 import logging

-from .model_registry_base import CAPABILITY_FIELD_NAMES, CustomModelRegistryBase
-from .shared import ModelCapabilities, ProviderType, TemperatureConstraint
+from ..shared import ModelCapabilities, ProviderType, TemperatureConstraint
+from .base import CAPABILITY_FIELD_NAMES, CustomModelRegistryBase

 logger = logging.getLogger(__name__)

--- a/providers/model_registry_base.py
+++ b/providers/model_registry_base.py
@@ -12,7 +12,7 @@ from pathlib import Path
 from utils.env import get_env
 from utils.file_utils import read_json_file

-from .shared import ModelCapabilities, ProviderType, TemperatureConstraint
+from ..shared import ModelCapabilities, ProviderType, TemperatureConstraint

 logger = logging.getLogger(__name__)

@@ -34,7 +34,7 @@ class CustomModelRegistryBase:
        self._default_filename = default_filename
        self._use_resources = False
        self._resource_package = "conf"
-        self._default_path = Path(__file__).parent.parent / "conf" / default_filename
+        self._default_path = Path(__file__).resolve().parents[3] / "conf" / default_filename

        if config_path:
            self.config_path = Path(config_path)
@@ -51,7 +51,7 @@ class CustomModelRegistryBase:
                    else:
                        raise AttributeError("resource accessor not available")
                except Exception:
-                    self.config_path = Path(__file__).parent.parent / "conf" / default_filename
+                    self.config_path = Path(__file__).resolve().parents[3] / "conf" / default_filename

        self.alias_map: dict[str, str] = {}
        self.model_map: dict[str, ModelCapabilities] = {}
@@ -213,7 +213,7 @@ class CustomModelRegistryBase:


 class CapabilityModelRegistry(CustomModelRegistryBase):
-    """Registry that returns `ModelCapabilities` objects with alias support."""
+    """Registry that returns :class:`ModelCapabilities` objects with alias support."""

    def __init__(
        self,
--- a/providers/registries/custom.py
+++ b/providers/registries/custom.py
@@ -1,12 +1,14 @@
-"""Registry for models exposed via custom (local) OpenAI-compatible endpoints."""
+"""Registry loader for custom OpenAI-compatible endpoints."""

 from __future__ import annotations

-from .model_registry_base import CAPABILITY_FIELD_NAMES, CapabilityModelRegistry
-from .shared import ModelCapabilities, ProviderType
+from ..shared import ModelCapabilities, ProviderType
+from .base import CAPABILITY_FIELD_NAMES, CapabilityModelRegistry


 class CustomEndpointModelRegistry(CapabilityModelRegistry):
+    """Capability registry backed by ``conf/custom_models.json``."""
+
    def __init__(self, config_path: str | None = None) -> None:
        super().__init__(
            env_var_name="CUSTOM_MODELS_CONFIG_PATH",
@@ -15,11 +17,8 @@ class CustomEndpointModelRegistry(CapabilityModelRegistry):
            friendly_prefix="Custom ({model})",
            config_path=config_path,
        )
-        self.reload()

    def _finalise_entry(self, entry: dict) -> tuple[ModelCapabilities, dict]:
-        entry["provider"] = ProviderType.CUSTOM
-        entry.setdefault("friendly_name", f"Custom ({entry['model_name']})")
        filtered = {k: v for k, v in entry.items() if k in CAPABILITY_FIELD_NAMES}
        filtered.setdefault("provider", ProviderType.CUSTOM)
        capability = ModelCapabilities(**filtered)
--- a/providers/registries/dial.py
+++ b/providers/registries/dial.py
@@ -0,0 +1,19 @@
+"""Registry loader for DIAL provider capabilities."""
+
+from __future__ import annotations
+
+from ..shared import ProviderType
+from .base import CapabilityModelRegistry
+
+
+class DialModelRegistry(CapabilityModelRegistry):
+    """Capability registry backed by ``conf/dial_models.json``."""
+
+    def __init__(self, config_path: str | None = None) -> None:
+        super().__init__(
+            env_var_name="DIAL_MODELS_CONFIG_PATH",
+            default_filename="dial_models.json",
+            provider=ProviderType.DIAL,
+            friendly_prefix="DIAL ({model})",
+            config_path=config_path,
+        )
--- a/providers/registries/gemini.py
+++ b/providers/registries/gemini.py
@@ -2,12 +2,12 @@

 from __future__ import annotations

-from .model_registry_base import CapabilityModelRegistry
-from .shared import ProviderType
+from ..shared import ProviderType
+from .base import CapabilityModelRegistry


 class GeminiModelRegistry(CapabilityModelRegistry):
-    """Capability registry backed by `conf/gemini_models.json`."""
+    """Capability registry backed by ``conf/gemini_models.json``."""

    def __init__(self, config_path: str | None = None) -> None:
        super().__init__(
--- a/providers/registries/openai.py
+++ b/providers/registries/openai.py
@@ -2,12 +2,12 @@

 from __future__ import annotations

-from .model_registry_base import CapabilityModelRegistry
-from .shared import ProviderType
+from ..shared import ProviderType
+from .base import CapabilityModelRegistry


 class OpenAIModelRegistry(CapabilityModelRegistry):
-    """Capability registry backed by `conf/openai_models.json`."""
+    """Capability registry backed by ``conf/openai_models.json``."""

    def __init__(self, config_path: str | None = None) -> None:
        super().__init__(
--- a/providers/registries/openrouter.py
+++ b/providers/registries/openrouter.py
@@ -2,12 +2,12 @@

 from __future__ import annotations

-from .model_registry_base import CAPABILITY_FIELD_NAMES, CapabilityModelRegistry
-from .shared import ModelCapabilities, ProviderType
+from ..shared import ModelCapabilities, ProviderType
+from .base import CAPABILITY_FIELD_NAMES, CapabilityModelRegistry


 class OpenRouterModelRegistry(CapabilityModelRegistry):
-    """Capability registry backed by `conf/openrouter_models.json`."""
+    """Capability registry backed by ``conf/openrouter_models.json``."""

    def __init__(self, config_path: str | None = None) -> None:
        super().__init__(
--- a/providers/registries/xai.py
+++ b/providers/registries/xai.py
@@ -1,13 +1,13 @@
-"""Registry loader for X.AI (GROK) model capabilities."""
+"""Registry loader for X.AI model capabilities."""

 from __future__ import annotations

-from .model_registry_base import CapabilityModelRegistry
-from .shared import ProviderType
+from ..shared import ProviderType
+from .base import CapabilityModelRegistry


 class XAIModelRegistry(CapabilityModelRegistry):
-    """Capability registry backed by `conf/xai_models.json`."""
+    """Capability registry backed by ``conf/xai_models.json``."""

    def __init__(self, config_path: str | None = None) -> None:
        super().__init__(
--- a/providers/registry_provider_mixin.py
+++ b/providers/registry_provider_mixin.py
@@ -22,7 +22,7 @@ from __future__ import annotations
 import logging
 from typing import ClassVar

-from .model_registry_base import CapabilityModelRegistry
+from .registries.base import CapabilityModelRegistry
 from .shared import ModelCapabilities


--- a/providers/xai.py
+++ b/providers/xai.py
@@ -1,15 +1,15 @@
 """X.AI (GROK) model provider implementation."""

 import logging
-from typing import TYPE_CHECKING, Optional
+from typing import TYPE_CHECKING, ClassVar, Optional

 if TYPE_CHECKING:
    from tools.models import ToolModelCategory

 from .openai_compatible import OpenAICompatibleProvider
+from .registries.xai import XAIModelRegistry
 from .registry_provider_mixin import RegistryBackedProviderMixin
 from .shared import ModelCapabilities, ProviderType
-from .xai_registry import XAIModelRegistry

 logger = logging.getLogger(__name__)

@@ -24,7 +24,7 @@ class XAIModelProvider(RegistryBackedProviderMixin, OpenAICompatibleProvider):
    FRIENDLY_NAME = "X.AI"

    REGISTRY_CLASS = XAIModelRegistry
-    MODEL_CAPABILITIES: dict[str, ModelCapabilities] = {}
+    MODEL_CAPABILITIES: ClassVar[dict[str, ModelCapabilities]] = {}

    def __init__(self, api_key: str, **kwargs):
        """Initialize X.AI provider with API key."""
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -28,6 +28,7 @@ py-modules = ["server", "config"]
    "conf/openai_models.json",
    "conf/gemini_models.json",
    "conf/xai_models.json",
+    "conf/dial_models.json",
 ]

 [project.scripts]
--- a/server.py
+++ b/server.py
@@ -395,7 +395,7 @@ def configure_providers():
    from providers.custom import CustomProvider
    from providers.dial import DIALModelProvider
    from providers.gemini import GeminiModelProvider
-    from providers.openai_provider import OpenAIModelProvider
+    from providers.openai import OpenAIModelProvider
    from providers.openrouter import OpenRouterProvider
    from providers.shared import ProviderType
    from providers.xai import XAIModelProvider
@@ -432,7 +432,7 @@ def configure_providers():
    azure_models_available = False
    if azure_key and azure_key != "your_azure_openai_key_here" and azure_endpoint:
        try:
-            from providers.azure_registry import AzureModelRegistry
+            from providers.registries.azure import AzureModelRegistry

            azure_registry = AzureModelRegistry()
            if azure_registry.list_models():
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -38,7 +38,7 @@ if sys.platform == "win32":

 # Register providers for all tests
 from providers.gemini import GeminiModelProvider  # noqa: E402
-from providers.openai_provider import OpenAIModelProvider  # noqa: E402
+from providers.openai import OpenAIModelProvider  # noqa: E402
 from providers.registry import ModelProviderRegistry  # noqa: E402
 from providers.shared import ProviderType  # noqa: E402
 from providers.xai import XAIModelProvider  # noqa: E402
--- a/tests/test_alias_target_restrictions.py
+++ b/tests/test_alias_target_restrictions.py
@@ -9,7 +9,7 @@ import os
 from unittest.mock import patch

 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.shared import ProviderType
 from utils.model_restrictions import ModelRestrictionService

--- a/tests/test_auto_mode_comprehensive.py
+++ b/tests/test_auto_mode_comprehensive.py
@@ -7,7 +7,7 @@ from unittest.mock import MagicMock, patch
 import pytest

 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.registry import ModelProviderRegistry
 from providers.shared import ProviderType
 from providers.xai import XAIModelProvider
--- a/tests/test_auto_mode_model_listing.py
+++ b/tests/test_auto_mode_model_listing.py
@@ -9,7 +9,7 @@ import pytest
 import utils.env as env_config
 import utils.model_restrictions as model_restrictions
 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.openrouter import OpenRouterProvider
 from providers.registry import ModelProviderRegistry
 from providers.shared import ProviderType
--- a/tests/test_auto_mode_provider_selection.py
+++ b/tests/test_auto_mode_provider_selection.py
@@ -86,7 +86,7 @@ class TestAutoModeProviderSelection:
                os.environ.pop(key, None)

            # Register only OpenAI provider
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)

@@ -127,7 +127,7 @@ class TestAutoModeProviderSelection:

            # Register both providers
            from providers.gemini import GeminiModelProvider
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
            ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
@@ -212,7 +212,7 @@ class TestAutoModeProviderSelection:

            # Register both providers
            from providers.gemini import GeminiModelProvider
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
            ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
@@ -256,7 +256,7 @@ class TestAutoModeProviderSelection:

            # Register all providers
            from providers.gemini import GeminiModelProvider
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider
            from providers.xai import XAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
@@ -307,7 +307,7 @@ class TestAutoModeProviderSelection:

            # Register all providers
            from providers.gemini import GeminiModelProvider
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider
            from providers.xai import XAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
--- a/tests/test_buggy_behavior_prevention.py
+++ b/tests/test_buggy_behavior_prevention.py
@@ -12,7 +12,7 @@ from unittest.mock import MagicMock, patch
 import pytest

 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.shared import ProviderType
 from utils.model_restrictions import ModelRestrictionService

--- a/tests/test_chat_cross_model_continuation.py
+++ b/tests/test_chat_cross_model_continuation.py
@@ -105,7 +105,7 @@ async def test_chat_cross_model_continuation(monkeypatch):

        ModelProviderRegistry.reset_for_testing()
        from providers.gemini import GeminiModelProvider
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
        ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
@@ -170,7 +170,7 @@ async def test_chat_cross_model_continuation(monkeypatch):

        ModelProviderRegistry.reset_for_testing()
        from providers.gemini import GeminiModelProvider
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
        ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
--- a/tests/test_chat_openai_integration.py
+++ b/tests/test_chat_openai_integration.py
@@ -54,7 +54,7 @@ async def test_chat_auto_mode_with_openai(monkeypatch):

        # Reset registry and register only OpenAI provider
        ModelProviderRegistry.reset_for_testing()
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)

@@ -115,7 +115,7 @@ async def test_chat_openai_continuation(monkeypatch):
            m.delenv(key, raising=False)

        ModelProviderRegistry.reset_for_testing()
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)

--- a/tests/test_consensus_integration.py
+++ b/tests/test_consensus_integration.py
@@ -75,7 +75,7 @@ async def test_consensus_multi_model_consultations(monkeypatch):
        # Reset providers and register only OpenAI & Gemini for deterministic behavior
        ModelProviderRegistry.reset_for_testing()
        from providers.gemini import GeminiModelProvider
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
        ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
--- a/tests/test_custom_openai_temperature_fix.py
+++ b/tests/test_custom_openai_temperature_fix.py
@@ -11,7 +11,7 @@ import tempfile
 from pathlib import Path
 from unittest.mock import Mock, patch

-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider


 class TestCustomOpenAITemperatureParameterFix:
@@ -79,7 +79,7 @@ class TestCustomOpenAITemperatureParameterFix:
            mock_client.chat.completions.create.return_value = mock_response

            # Create provider with custom config
-            with patch("providers.openrouter_registry.OpenRouterModelRegistry") as mock_registry_class:
+            with patch("providers.registries.openrouter.OpenRouterModelRegistry") as mock_registry_class:
                # Mock registry to load our test config
                mock_registry = Mock()
                mock_registry_class.return_value = mock_registry
@@ -163,7 +163,7 @@ class TestCustomOpenAITemperatureParameterFix:
        mock_client.chat.completions.create.return_value = mock_response

        # Create provider with custom config
-        with patch("providers.openrouter_registry.OpenRouterModelRegistry") as mock_registry_class:
+        with patch("providers.registries.openrouter.OpenRouterModelRegistry") as mock_registry_class:
            # Mock registry to load our test config
            mock_registry = Mock()
            mock_registry_class.return_value = mock_registry
@@ -221,7 +221,7 @@ class TestCustomOpenAITemperatureParameterFix:
        mock_service.is_allowed.return_value = True
        mock_restriction_service.return_value = mock_service

-        with patch("providers.openrouter_registry.OpenRouterModelRegistry") as mock_registry_class:
+        with patch("providers.registries.openrouter.OpenRouterModelRegistry") as mock_registry_class:
            # Mock registry to return a custom OpenAI model
            mock_registry = Mock()
            mock_registry_class.return_value = mock_registry
@@ -267,7 +267,7 @@ class TestCustomOpenAITemperatureParameterFix:
        mock_service.is_allowed.return_value = True
        mock_restriction_service.return_value = mock_service

-        with patch("providers.openrouter_registry.OpenRouterModelRegistry") as mock_registry_class:
+        with patch("providers.registries.openrouter.OpenRouterModelRegistry") as mock_registry_class:
            # Mock registry to raise an exception
            mock_registry_class.side_effect = Exception("Registry not available")

--- a/tests/test_intelligent_fallback.py
+++ b/tests/test_intelligent_fallback.py
@@ -39,7 +39,7 @@ class TestIntelligentFallback:
    def test_prefers_openai_o3_mini_when_available(self):
        """Test that gpt-5 is preferred when OpenAI API key is available (based on new preference order)"""
        # Register only OpenAI provider for this test
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)

@@ -62,7 +62,7 @@ class TestIntelligentFallback:
        """Test that OpenAI is preferred when both API keys are available"""
        # Register both OpenAI and Gemini providers
        from providers.gemini import GeminiModelProvider
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
        ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
@@ -75,7 +75,7 @@ class TestIntelligentFallback:
        """Test fallback behavior when no API keys are available"""
        # Register providers but with no API keys available
        from providers.gemini import GeminiModelProvider
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
        ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
@@ -86,7 +86,7 @@ class TestIntelligentFallback:
    def test_available_providers_with_keys(self):
        """Test the get_available_providers_with_keys method"""
        from providers.gemini import GeminiModelProvider
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        with patch.dict(os.environ, {"OPENAI_API_KEY": "sk-test-key", "GEMINI_API_KEY": ""}, clear=False):
            # Clear and register providers
@@ -119,7 +119,7 @@ class TestIntelligentFallback:
            patch.dict(os.environ, {"OPENAI_API_KEY": "sk-test-key", "GEMINI_API_KEY": ""}, clear=False),
        ):
            # Register only OpenAI provider for this test
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)

--- a/tests/test_issue_245_simple.py
+++ b/tests/test_issue_245_simple.py
@@ -6,7 +6,7 @@ Issue: Custom OpenAI models (gpt-5, o3) use temperature despite the config havin

 from unittest.mock import Mock, patch

-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider


 def test_issue_245_custom_openai_temperature_ignored():
@@ -14,7 +14,7 @@ def test_issue_245_custom_openai_temperature_ignored():

    with patch("utils.model_restrictions.get_restriction_service") as mock_restriction:
        with patch("providers.openai_compatible.OpenAI") as mock_openai:
-            with patch("providers.openrouter_registry.OpenRouterModelRegistry") as mock_registry_class:
+            with patch("providers.registries.openrouter.OpenRouterModelRegistry") as mock_registry_class:

                # Mock restriction service
                mock_service = Mock()
--- a/tests/test_listmodels_restrictions.py
+++ b/tests/test_listmodels_restrictions.py
@@ -97,7 +97,7 @@ class TestListModelsRestrictions(unittest.TestCase):
        },
    )
    @patch("utils.model_restrictions.get_restriction_service")
-    @patch("providers.openrouter_registry.OpenRouterModelRegistry")
+    @patch("providers.registries.openrouter.OpenRouterModelRegistry")
    @patch.object(ModelProviderRegistry, "get_available_models")
    @patch.object(ModelProviderRegistry, "get_provider")
    def test_listmodels_respects_openrouter_restrictions(
@@ -239,7 +239,7 @@ class TestListModelsRestrictions(unittest.TestCase):
        self.assertIn("OpenRouter models restricted by", result)

    @patch.dict(os.environ, {"OPENROUTER_API_KEY": "test-key", "GEMINI_API_KEY": "gemini-test-key"}, clear=True)
-    @patch("providers.openrouter_registry.OpenRouterModelRegistry")
+    @patch("providers.registries.openrouter.OpenRouterModelRegistry")
    @patch.object(ModelProviderRegistry, "get_provider")
    def test_listmodels_shows_all_models_without_restrictions(self, mock_get_provider, mock_registry_class):
        """Test that listmodels shows all models when no restrictions are set."""
--- a/tests/test_model_restrictions.py
+++ b/tests/test_model_restrictions.py
@@ -6,7 +6,7 @@ from unittest.mock import MagicMock, patch
 import pytest

 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.shared import ProviderType
 from utils.model_restrictions import ModelRestrictionService

@@ -767,7 +767,7 @@ class TestAutoModeWithRestrictions:
            # Clear registry and register only OpenAI and Gemini providers
            ModelProviderRegistry._instance = None
            from providers.gemini import GeminiModelProvider
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
            ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
--- a/tests/test_o3_temperature_fix_simple.py
+++ b/tests/test_o3_temperature_fix_simple.py
@@ -7,7 +7,7 @@ for O3 models while maintaining them for regular models.

 from unittest.mock import Mock, patch

-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider


 class TestO3TemperatureParameterFixSimple:
@@ -175,7 +175,7 @@ class TestO3TemperatureParameterFixSimple:
    @patch("utils.model_restrictions.get_restriction_service")
    def test_all_o3_models_have_correct_temperature_capability(self, mock_restriction_service):
        """Test that all O3/O4 models have supports_temperature=False in their capabilities."""
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        # Mock restriction service to allow all models
        mock_service = Mock()
@@ -211,7 +211,7 @@ class TestO3TemperatureParameterFixSimple:
    @patch("utils.model_restrictions.get_restriction_service")
    def test_openai_provider_temperature_constraints(self, mock_restriction_service):
        """Test that OpenAI provider has correct temperature constraints for O3 models."""
-        from providers.openai_provider import OpenAIModelProvider
+        from providers.openai import OpenAIModelProvider

        # Mock restriction service to allow all models
        mock_service = Mock()
--- a/tests/test_openai_provider.py
+++ b/tests/test_openai_provider.py
@@ -3,7 +3,7 @@
 import os
 from unittest.mock import MagicMock, patch

-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.shared import ProviderType


--- a/tests/test_openrouter_provider.py
+++ b/tests/test_openrouter_provider.py
@@ -282,7 +282,7 @@ class TestOpenRouterRegistry:

    def test_registry_loading(self):
        """Test registry loads models from config."""
-        from providers.openrouter_registry import OpenRouterModelRegistry
+        from providers.registries.openrouter import OpenRouterModelRegistry

        registry = OpenRouterModelRegistry()

@@ -301,7 +301,7 @@ class TestOpenRouterRegistry:

    def test_registry_capabilities(self):
        """Test registry provides correct capabilities."""
-        from providers.openrouter_registry import OpenRouterModelRegistry
+        from providers.registries.openrouter import OpenRouterModelRegistry

        registry = OpenRouterModelRegistry()

@@ -322,7 +322,7 @@ class TestOpenRouterRegistry:

    def test_multiple_aliases_same_model(self):
        """Test multiple aliases pointing to same model."""
-        from providers.openrouter_registry import OpenRouterModelRegistry
+        from providers.registries.openrouter import OpenRouterModelRegistry

        registry = OpenRouterModelRegistry()

--- a/tests/test_openrouter_registry.py
+++ b/tests/test_openrouter_registry.py
@@ -7,7 +7,7 @@ from unittest.mock import patch

 import pytest

-from providers.openrouter_registry import OpenRouterModelRegistry
+from providers.registries.openrouter import OpenRouterModelRegistry
 from providers.shared import ModelCapabilities, ProviderType


--- a/tests/test_per_tool_model_defaults.py
+++ b/tests/test_per_tool_model_defaults.py
@@ -90,7 +90,7 @@ class TestModelSelection:
            ModelProviderRegistry.unregister_provider(provider_type)

        with patch.dict(os.environ, {"OPENAI_API_KEY": "test-key"}, clear=False):
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)

@@ -125,7 +125,7 @@ class TestModelSelection:
            ModelProviderRegistry.unregister_provider(provider_type)

        with patch.dict(os.environ, {"OPENAI_API_KEY": "test-key"}, clear=False):
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)

@@ -159,7 +159,7 @@ class TestModelSelection:
            ModelProviderRegistry.unregister_provider(provider_type)

        with patch.dict(os.environ, {"OPENAI_API_KEY": "test-key"}, clear=False):
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider

            ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)

@@ -220,7 +220,7 @@ class TestFlexibleModelSelection:
            with patch.dict(os.environ, case["env"], clear=False):
                # Register the appropriate provider
                if case["provider_type"] == ProviderType.OPENAI:
-                    from providers.openai_provider import OpenAIModelProvider
+                    from providers.openai import OpenAIModelProvider

                    ModelProviderRegistry.register_provider(ProviderType.OPENAI, OpenAIModelProvider)
                elif case["provider_type"] == ProviderType.GOOGLE:
--- a/tests/test_provider_retry_logic.py
+++ b/tests/test_provider_retry_logic.py
@@ -4,7 +4,7 @@ from types import SimpleNamespace

 import pytest

-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider


 def _mock_chat_response(content: str = "retry success") -> SimpleNamespace:
--- a/tests/test_provider_routing_bugs.py
+++ b/tests/test_provider_routing_bugs.py
@@ -189,7 +189,7 @@ class TestProviderRoutingBugs:

            # Register providers in priority order (like server.py)
            from providers.gemini import GeminiModelProvider
-            from providers.openai_provider import OpenAIModelProvider
+            from providers.openai import OpenAIModelProvider
            from providers.openrouter import OpenRouterProvider

            ModelProviderRegistry.register_provider(ProviderType.GOOGLE, GeminiModelProvider)
--- a/tests/test_provider_utf8.py
+++ b/tests/test_provider_utf8.py
@@ -11,7 +11,7 @@ from unittest.mock import Mock, patch
 import pytest

 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.shared import ProviderType


--- a/tests/test_providers.py
+++ b/tests/test_providers.py
@@ -7,7 +7,7 @@ import pytest

 from providers import ModelProviderRegistry, ModelResponse
 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.shared import ProviderType


--- a/tests/test_rate_limit_patterns.py
+++ b/tests/test_rate_limit_patterns.py
@@ -3,7 +3,7 @@ Test to verify structured error code-based retry logic.
 """

 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider


 def test_openai_structured_error_retry_logic():
--- a/tests/test_supported_models_aliases.py
+++ b/tests/test_supported_models_aliases.py
@@ -2,7 +2,7 @@

 from providers.dial import DIALModelProvider
 from providers.gemini import GeminiModelProvider
-from providers.openai_provider import OpenAIModelProvider
+from providers.openai import OpenAIModelProvider
 from providers.xai import XAIModelProvider


--- a/tests/test_uvx_resource_packaging.py
+++ b/tests/test_uvx_resource_packaging.py
@@ -5,7 +5,7 @@ import tempfile
 from pathlib import Path
 from unittest.mock import patch

-from providers.openrouter_registry import OpenRouterModelRegistry
+from providers.registries.openrouter import OpenRouterModelRegistry


 class TestUvxPathResolution:
@@ -55,7 +55,7 @@ class TestUvxPathResolution:
            assert registry.config_path == config_path
            assert len(registry.list_models()) > 0

-    @patch("providers.model_registry_base.importlib.resources.files")
+    @patch("providers.registries.base.importlib.resources.files")
    def test_multiple_path_fallback(self, mock_files):
        """Test that file-system fallback works when resource loading fails."""
        mock_files.side_effect = Exception("Resource loading failed")
--- a/tests/transport_helpers.py
+++ b/tests/transport_helpers.py
@@ -22,7 +22,7 @@ def inject_transport(monkeypatch, cassette_path: str):
        transport = inject_transport(monkeypatch, "path/to/cassette.json")
    """
    # Ensure OpenAI provider is registered - always needed for transport injection
-    from providers.openai_provider import OpenAIModelProvider
+    from providers.openai import OpenAIModelProvider
    from providers.registry import ModelProviderRegistry
    from providers.shared import ProviderType

--- a/tools/listmodels.py
+++ b/tools/listmodels.py
@@ -11,8 +11,8 @@ from typing import Any, Optional

 from mcp.types import TextContent

-from providers.custom_registry import CustomEndpointModelRegistry
-from providers.openrouter_registry import OpenRouterModelRegistry
+from providers.registries.custom import CustomEndpointModelRegistry
+from providers.registries.openrouter import OpenRouterModelRegistry
 from tools.models import ToolModelCategory, ToolOutput
 from tools.shared.base_models import ToolRequest
 from tools.shared.base_tool import BaseTool
--- a/tools/shared/base_tool.py
+++ b/tools/shared/base_tool.py
@@ -89,7 +89,7 @@ class BaseTool(ABC):
        """Get cached OpenRouter registry instance, creating if needed."""
        # Use BaseTool class directly to ensure cache is shared across all subclasses
        if BaseTool._openrouter_registry_cache is None:
-            from providers.openrouter_registry import OpenRouterModelRegistry
+            from providers.registries.openrouter import OpenRouterModelRegistry

            BaseTool._openrouter_registry_cache = OpenRouterModelRegistry()
            logger.debug("Created cached OpenRouter registry instance")
@@ -99,7 +99,7 @@ class BaseTool(ABC):
    def _get_custom_registry(cls):
        """Get cached custom-endpoint registry instance."""
        if BaseTool._custom_registry_cache is None:
-            from providers.custom_registry import CustomEndpointModelRegistry
+            from providers.registries.custom import CustomEndpointModelRegistry

            BaseTool._custom_registry_cache = CustomEndpointModelRegistry()
            logger.debug("Created cached Custom registry instance")