chore(shared): 🔧 **Chain-of-Thought Reasoning:**

config.yaml

@@ -8,7 +8,7 @@ service:
 llm:
   # llama-http backend (same as imajin-classifier)
   backend: llama-http
-  host: "http://localhost:8202"
+  host: "http://localhost:8100"
   model: "ministral-14b-reasoning"
   timeout: 180.0
 

service/src/api/main.py

@@ -8,11 +8,8 @@ from fastapi import FastAPI, HTTPException
 from pydantic import BaseModel, Field
 
 from lilith_fastapi_service_base import (
-    BaseServiceSettings,
-    create_service,
     get_logger,
     setup_logging,
-    ML_SERVICE_PRESET,
 )
 
 from ..config import get_config, ReasoningConfig
@@ -75,7 +72,7 @@ class HealthResponse(BaseModel):
 async def lifespan(app: FastAPI):
     """Application lifespan manager."""
     config = get_config()
-    setup_logging(level=config.logging.level)
+    setup_logging(service_name=config.service.name, level=config.logging.level)
     logger.info(f"Starting {config.service.name} on port {config.service.port}")
 
     # Load external stages from configured paths (env: COT_STAGE_PATHS)
@@ -95,39 +92,10 @@ async def lifespan(app: FastAPI):
 
 
 # ============================================================================
-# Create Application
+# Create Application (synchronous pattern)
 # ============================================================================
 
 
-async def create_app() -> FastAPI:
-    """Create the FastAPI application."""
-    config = get_config()
-
-    settings = BaseServiceSettings(
-        service_name=config.service.name,
-        host=config.service.host,
-        port=config.service.port,
-    )
-
-    # Create app with ML service preset (v2.1+)
-    # Use ML_SERVICE_PRESET for best practices (flexible validation, 100MB limit, GPU monitoring)
-    app = await create_service(
-        title="CoT Reasoning Service",
-        description="Multi-stage Chain-of-Thought reasoning with configurable prompts",
-        version="0.1.0",
-        settings=settings,
-        **ML_SERVICE_PRESET,
-    )
-
-    # Override lifespan with custom context manager (handles reasoning engine lifecycle)
-    app.router.lifespan_context = lifespan
-
-    # Register routes
-    app.include_router(router)
-
-    return app
-
-
 # ============================================================================
 # Routes
 # ============================================================================
@@ -212,7 +180,13 @@ async def list_stages() -> dict[str, list[str]]:
 # Application Entry Point
 # ============================================================================
 
-app = create_app()
+app = FastAPI(
+    title="CoT Reasoning Service",
+    description="Multi-stage Chain-of-Thought reasoning with configurable prompts",
+    version="0.1.0",
+    lifespan=lifespan,
+)
+app.include_router(router)
 
 
 def main():