Fix langchain deprecation + lazy init RAGExplainer

rag/explainer.py

@@ -1,4 +1,4 @@
-from langchain.chat_models import ChatOpenAI
+from langchain_openai import ChatOpenAI
 from langchain.prompts import ChatPromptTemplate
 from langchain.schema import HumanMessage, SystemMessage
 from typing import Dict, List
@@ -12,10 +12,10 @@ class RAGExplainer:
         llm_kwargs = {
             "model": "gpt-4.1-nano",
             "temperature": 0.3,
-            "openai_api_key": settings.OPENAI_API_KEY
+            "api_key": settings.OPENAI_API_KEY
         }
         if settings.OPENAI_BASE_URL:
-            llm_kwargs["openai_api_base"] = settings.OPENAI_BASE_URL
+            llm_kwargs["base_url"] = settings.OPENAI_BASE_URL
         
         self.llm = ChatOpenAI(**llm_kwargs)
     

routes/analyze.py

@@ -18,7 +18,14 @@ router = APIRouter(prefix="/analyze", tags=["Resume Analysis"])
 # Initialize services
 embedding_model = EmbeddingModel(settings.EMBEDDING_MODEL)
 ranking_engine = RankingEngine()
-rag_explainer = RAGExplainer()
+rag_explainer = None  # Lazy init to avoid startup crash if no API key
+
+
+def get_rag_explainer():
+    global rag_explainer
+    if rag_explainer is None:
+        rag_explainer = RAGExplainer()
+    return rag_explainer
 
 
 class AnalysisResponse(BaseModel):
@@ -107,7 +114,7 @@ async def analyze_resume(
         )
         
         # Generate AI explanation and suggestions
-        explanation = rag_explainer.generate_explanation(
+        explanation = get_rag_explainer().generate_explanation(
             resume_text=resume_text,
             job_description=job_text,
             ranking_result=ranking_result,
@@ -184,7 +191,7 @@ async def analyze_resume_text(
         )
         
         # Generate AI explanation
-        explanation = rag_explainer.generate_explanation(
+        explanation = get_rag_explainer().generate_explanation(
             resume_text=resume_text,
             job_description=job_description,
             ranking_result=ranking_result,