Merge branch 'NVIDIA:main' into main

Author: raosukrit67

README.md

@@ -48,7 +48,7 @@ This repository provides a sample code to demonstrate how you can use the log an
 
 # Software Components
 NVIDIA NIM Microservices
-- NIM of meta/llama-3.1-70b-instruct
+- NIM of nvidia/llama-3.3-nemotron-super-49b-v1.5
 - Retriever Models
 - NIM of nvidia/llama-3_2-nv-embedqa-1b-v2
 - NIM of nvidia/llama-3_2-nv-rerankqa-1b-v2

community/log_analysis_multi_agent_rag/README.md

@@ -11,6 +11,12 @@ def decide_to_generate(state):
         print("ASSESS GRADED DOCUMENTS")
         state["question"]
         filtered_documents = state["documents"]
+        transform_count = state.get("transform_count", 0)
+
+        # If we've transformed too many times, force generation
+        if transform_count >= 2:
+            print("DECISION: MAX TRANSFORMS REACHED, FORCING GENERATION")
+            return "generate"
 
         if not filtered_documents:
             print(
@@ -33,10 +39,21 @@ def grade_generation_vs_documents_and_question(state):
         generation = state["generation"]
 
         print("GRADE GENERATED vs QUESTION")
-        score = automation.answer_grader.invoke({"question": question, "generation": generation})
-        grade = score.binary_score
-        if grade == "yes":
-            print("DECISION: GENERATION ADDRESSES QUESTION")
-            return "useful"
-        print("DECISION: GENERATION DOES NOT ADDRESS QUESTION")
-        return "not useful"
+        try:
+            score_text = automation.answer_grader.invoke({"question": question, "generation": generation})
+            if "yes" in score_text.lower():
+                print("DECISION: GENERATION ADDRESSES QUESTION")
+                return "useful"
+            else:
+                # Check if we've transformed too many times
+                transform_count = state.get("transform_count", 0)
+                if transform_count >= 2:
+                    print("DECISION: MAX TRANSFORMS REACHED, ACCEPTING GENERATION")
+                    return "useful"
+                else:
+                    print("DECISION: GENERATION DOES NOT ADDRESS QUESTION")
+                    return "not useful"
+        except:
+            # If grading fails, assume generation is useful to avoid infinite loops
+            print("DECISION: GRADING FAILED, ACCEPTING GENERATION")
+            return "useful"

community/log_analysis_multi_agent_rag/graphedges.py

@@ -18,13 +18,13 @@ def __init__(self, file_path, api_key):
         self.hybrid_retriever = self.create_hybrid_retriever()
 
     def initialize_nvidia_components(self):
-        embeddings =NVIDIAEmbeddings(model="nvidia/llama-3.2-nv-embedqa-1b-v2", truncate="NONE")
+        embeddings =NVIDIAEmbeddings(model="nvidia/llama-3.2-nv-embedqa-1b-v2", truncate="END")
         return  embeddings
 
     def load_and_split_documents(self):
         loader = TextLoader(self.file_path)
         docs = loader.load()
-        text_splitter = RecursiveCharacterTextSplitter(chunk_size=5000, chunk_overlap=600)
+        text_splitter = RecursiveCharacterTextSplitter(chunk_size=20000, chunk_overlap=10000)
         doc_splits = text_splitter.split_documents(docs)
         return doc_splits
 
@@ -35,7 +35,7 @@ def create_retrievers(self):
         return bm25_retriever, faiss_retriever
 
     def create_hybrid_retriever(self):
-        hybrid_retriever = EnsembleRetriever(retrievers=[self.bm25_retriever, self.faiss_retriever], weights=[0.7, 0.3])
+        hybrid_retriever = EnsembleRetriever(retrievers=[self.bm25_retriever, self.faiss_retriever], weights=[0.5, 0.5])
         return hybrid_retriever
 
     def get_retriever(self):