feat: Enhance usage tracking and model capabilities

This commit introduces several improvements to the usage tracking and model capabilities systems.

- **Refactor Usage Tracking:** The usage tracking statistics endpoint now returns a UsageStatsResponse Pydantic model instead of a raw dictionary. This improves type safety, ensures a consistent data structure, and enables better validation. The UsageController, IUsageTrackingService, and UsageTrackingService have been updated accordingly.

- **Improve Model Capabilities:** The default reasoning and execution phase parameters now include 
easoning_effort. This ensures that even unknown backends have a default reasoning effort setting, improving the robustness of the reasoning system.

- **Update Tests:** All relevant unit tests for usage tracking have been updated to align with the new UsageStatsResponse model and to ensure comprehensive coverage of the changes.

Author: matdev83

data/test_suite_state.json

@@ -1,4 +1,4 @@
 {
-  "test_count": 5119,
-  "last_updated": "1762168167.0802596"
+  "test_count": 5046,
+  "last_updated": "1762217835.0"
 }

src/connectors/utils/model_capabilities.py

@@ -27,18 +27,18 @@
 REASONING_TAG_FORMAT: dict[str, tuple[str, str]] = {
     "openai": ("<thinking>", "</thinking>"),
     "anthropic": ("<thinking>", "</thinking>"),
-    "qwen": ("<thinking>", "</thinking>"),
-    "qwen-oauth": ("<thinking>", "</thinking>"),
-    "deepseek": ("<think>", "</think>"),  # DeepSeek uses <think> natively
-    "minimax": ("<think>", "</think>"),
-    "gemini": ("<thinking>", "</thinking>"),
-    "gemini-oauth-plan": ("<thinking>", "</thinking>"),
-    "gemini-oauth-free": ("<thinking>", "</thinking>"),
-    "gemini-cli-acp": ("<thinking>", "</thinking>"),
-    "gemini-cli-cloud-project": ("<thinking>", "</thinking>"),
-    # Default for others
-    "_default": ("<reasoning>", "</reasoning>"),
-}
+    "qwen": ("<thinking>", "</thinking>"),
+    "qwen-oauth": ("<thinking>", "</thinking>"),
+    "deepseek": ("<think>", "</think>"),  # DeepSeek uses <think> natively
+    "minimax": ("<think>", "</think>"),
+    "gemini": ("<thinking>", "</thinking>"),
+    "gemini-oauth-plan": ("<thinking>", "</thinking>"),
+    "gemini-oauth-free": ("<thinking>", "</thinking>"),
+    "gemini-cli-acp": ("<thinking>", "</thinking>"),
+    "gemini-cli-cloud-project": ("<thinking>", "</thinking>"),
+    # Default for others
+    "_default": ("<reasoning>", "</reasoning>"),
+}
 
 # Reasoning phase parameters - maximize reasoning quality
 REASONING_PHASE_PARAMS: dict[str, dict[str, Any]] = {
@@ -63,6 +63,7 @@
     "_default": {
         # Generic reasoning parameters
         "temperature": 0.7,  # Balanced for reasoning
+        "reasoning_effort": "high",  # Ensure strong reasoning on unknown backends
     },
 }
 
@@ -87,6 +88,7 @@
     "_default": {
         # Generic execution parameters
         "temperature": 0.5,  # Lower for consistency
+        "reasoning_effort": "low",  # Minimize reasoning for unknown backends
     },
 }
 

src/core/app/controllers/usage_controller.py

@@ -5,12 +5,13 @@
 from __future__ import annotations
 
 import logging
-from typing import Any
+from typing import Any, cast
 
 from fastapi import APIRouter, Depends, Query
 
 from src.core.di.services import get_or_build_service_provider
 from src.core.domain.usage_data import UsageData
+from src.core.domain.usage_stats import UsageStatsResponse
 from src.core.interfaces.usage_tracking_interface import IUsageTrackingService
 
 logger = logging.getLogger(__name__)
@@ -31,21 +32,21 @@ def __init__(self, usage_service: IUsageTrackingService | None = None) -> None:
 
     async def get_usage_stats(
         self, project: str | None = None, days: int = 30
-    ) -> dict[str, Any]:
+    ) -> UsageStatsResponse:
         """Get usage statistics.
 
         Args:
             project: Optional project filter
             days: Number of days to include in stats
 
         Returns:
-            Usage statistics dictionary
+            Usage statistics response model
         """
         if not self.usage_service:
-            return {"error": "Usage tracking service not available"}
+            raise RuntimeError("Usage tracking service not available")
 
         result = await self.usage_service.get_usage_stats(project=project, days=days)
-        return result  # type: ignore[no-any-return]
+        return result
 
     async def get_recent_usage(
         self, session_id: str | None = None, limit: int = 100
@@ -68,12 +69,12 @@ async def get_recent_usage(
         return result  # type: ignore[no-any-return]
 
 
-@router.get("/stats", response_model=dict[str, Any])
+@router.get("/stats", response_model=UsageStatsResponse)
 async def get_usage_stats(
     project: str | None = Query(None, description="Filter by project name"),
     days: int = Query(30, description="Number of days to include in stats"),
     service_provider: Any = Depends(get_or_build_service_provider),
-) -> dict[str, Any]:
+) -> UsageStatsResponse:
     """Get usage statistics.
 
     Args:
@@ -84,9 +85,12 @@ async def get_usage_stats(
     Returns:
         Usage statistics dictionary
     """
-    usage_service = service_provider.get_required_service(IUsageTrackingService)
+    usage_service = cast(
+        IUsageTrackingService,
+        service_provider.get_required_service(IUsageTrackingService),
+    )
     result = await usage_service.get_usage_stats(project=project, days=days)
-    return result  # type: ignore[no-any-return]
+    return result
 
 
 @router.get("/recent", response_model=list[UsageData])

src/core/interfaces/usage_tracking_interface.py

@@ -8,6 +8,7 @@
 from typing import Any
 
 from src.core.domain.usage_data import UsageData
+from src.core.domain.usage_stats import UsageStatsResponse
 
 
 class IUsageTrackingService(abc.ABC):
@@ -44,7 +45,7 @@ async def track_request(
     @abc.abstractmethod
     async def get_usage_stats(
         self, project: str | None = None, days: int = 30
-    ) -> dict[str, Any]:
+    ) -> UsageStatsResponse:
         pass
 
     @abc.abstractmethod

src/core/services/usage_tracking_service.py

@@ -274,23 +274,27 @@ def _parse_billing_cost(value: Any) -> float | None:
 
     async def get_usage_stats(
         self, project: str | None = None, days: int = 30
-    ) -> dict[str, Any]:
-        """Get usage statistics.
+    ) -> UsageStatsResponse:
+        """Get usage statistics aggregated by model.
 
         Args:
             project: Optional project filter
             days: Number of days to include in stats
 
         Returns:
-            Usage statistics dictionary
+            A :class:`UsageStatsResponse` with aggregated usage metrics.
         """
         if days <= 0:
             logger.warning(
                 "Received non-positive days=%s when requesting usage stats; "
                 "falling back to complete history.",
                 days,
             )
-            return await self._repository.get_stats(project)
+            raw_stats = await self._repository.get_stats(project)
+            stats_response = UsageStatsResponse()
+            for model_name, payload in raw_stats.items():
+                stats_response[model_name] = payload
+            return stats_response
 
         usage_records = await self._repository.get_all()
         cutoff = datetime.now(timezone.utc) - timedelta(days=days)
@@ -329,7 +333,7 @@ async def get_usage_stats(
                 requests=current_stats.requests + 1,
             )
 
-        return stats.model_dump()
+        return stats
 
     async def get_recent_usage(
         self, session_id: str | None = None, limit: int = 100