Modified the prompt templates

backend/chains/router.py

@@ -85,7 +85,6 @@ from langchain_chroma import Chroma
 def init_vector_store():
     embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
 
-    # Attempt to pull prebuilt vector store
     try:
         path = fetch_vector_store_if_needed(REPO_URL, BRANCH, REPO_VECTOR_SUBFOLDER, VECTOR_STORE_PATH)
         print(f"📂 Using vector store from {path}")
@@ -125,7 +124,12 @@ from langchain_community.chat_message_histories import ChatMessageHistory
 
 from backend.schemas.api_models import ChatInput, ChatOutput
 from backend.prompts.templates import (
-    rag_prompt, quiz_generator_prompt, flashcard_generator_prompt, condense_question_prompt
+    rag_prompt,
+    quiz_generator_prompt,
+    flashcard_generator_prompt,
+    explanation_prompt,
+    definition_prompt,
+    condense_question_prompt
 )
 
 def format_docs(docs): return "\n---\n".join(doc.page_content for doc in docs)
@@ -144,34 +148,35 @@ def get_memory_for_session(session_id: str):
 # Retriever
 def EducationalRetriever(): return vector_store.as_retriever(search_kwargs={"k": 5})
 
+# -------------------------
 # Conversation chain
+# -------------------------
 def AdaptiveConversationChain():
     retriever = EducationalRetriever()
     condense = (
-        RunnableLambda(lambda x: {"question": x["input"], "chat_history": get_buffer_string(x["chat_history"])})
+        RunnableLambda(lambda x: {"question": x["input"], "chat_history": get_buffer_string(x["chat_history"])} )
         | condense_question_prompt | openai_llm | StrOutputParser()
     )
     return (
         RunnablePassthrough.assign(standalone_question=condense)
         .assign(context=(RunnableLambda(lambda x: x["standalone_question"]) | retriever))
         | RunnableParallel(
-            answer=(
-                RunnableLambda(lambda x: {
-                    "context": format_docs(x["context"]),
-                    "question": x["input"],
-                    "subject": x.get("subject", "the topic"),
-                    "difficulty_level": x.get("difficulty_level", "beginner"),
-                }) | rag_prompt | finetuned_llm | StrOutputParser()
-            ),
+            answer=(RunnableLambda(lambda x: {
+                "context": format_docs(x["context"]),
+                "question": x["input"],
+                "subject": x.get("subject", "the topic"),
+                "difficulty_level": x.get("difficulty_level", "beginner"),
+            }) | rag_prompt | finetuned_llm | StrOutputParser()),
             sources=RunnableLambda(lambda x: get_sources_from_docs(x["context"])),
         )
     )
 
-# Content generation (quiz + flashcards)
-def ContentGenerator():
-    retriever = EducationalRetriever()
-    QuizChain = (
-        RunnablePassthrough.assign(context=(RunnableLambda(lambda x: x["input"]) | retriever))
+# -------------------------
+# Generalized Content Chains
+# -------------------------
+def QuizChain():
+    return (
+        RunnablePassthrough.assign(context=(RunnableLambda(lambda x: x["input"]) | EducationalRetriever()))
         | RunnableParallel(
             answer=(RunnableLambda(lambda x: {
                 "context": format_docs(x["context"]),
@@ -181,8 +186,10 @@ def ContentGenerator():
             sources=RunnableLambda(lambda x: get_sources_from_docs(x["context"])),
         )
     )
-    FlashcardChain = (
-        RunnablePassthrough.assign(context=(RunnableLambda(lambda x: x["input"]) | retriever))
+
+def FlashcardChain():
+    return (
+        RunnablePassthrough.assign(context=(RunnableLambda(lambda x: x["input"]) | EducationalRetriever()))
         | RunnableParallel(
             answer=(RunnableLambda(lambda x: {
                 "context": format_docs(x["context"]),
@@ -191,29 +198,69 @@ def ContentGenerator():
             sources=RunnableLambda(lambda x: get_sources_from_docs(x["context"])),
         )
     )
+
+def ExplanationChain():
+    return (
+        RunnablePassthrough.assign(context=(RunnableLambda(lambda x: x["input"]) | EducationalRetriever()))
+        | RunnableParallel(
+            answer=(RunnableLambda(lambda x: {
+                "context": format_docs(x["context"]),
+                "topic": x.get("subject", "topic"),
+                "difficulty_level": x.get("difficulty_level", "beginner"),
+            }) | explanation_prompt | finetuned_llm | StrOutputParser()),
+            sources=RunnableLambda(lambda x: get_sources_from_docs(x["context"])),
+        )
+    )
+
+def DefinitionChain():
+    return (
+        RunnablePassthrough.assign(context=(RunnableLambda(lambda x: x["input"]) | EducationalRetriever()))
+        | RunnableParallel(
+            answer=(RunnableLambda(lambda x: {
+                "context": format_docs(x["context"]),
+                "term": x.get("subject", "term"),
+                "difficulty_level": x.get("difficulty_level", "beginner"),
+            }) | definition_prompt | finetuned_llm | StrOutputParser()),
+            sources=RunnableLambda(lambda x: get_sources_from_docs(x["context"])),
+        )
+    )
+
+# -------------------------
+# Dispatcher for content generation
+# -------------------------
+def ContentGenerator():
     return RunnableBranch(
-        (lambda x: x.get("request_type") == "quiz_generation", QuizChain),
-        (lambda x: x.get("request_type") == "flashcard_creation", FlashcardChain),
+        (lambda x: x.get("request_type") == "quiz_generation", QuizChain()),
+        (lambda x: x.get("request_type") == "flashcard_creation", FlashcardChain()),
+        (lambda x: x.get("request_type") == "explanation", ExplanationChain()),
+        (lambda x: x.get("request_type") == "definition", DefinitionChain()),
         RunnableLambda(lambda _: {"answer": "Unknown request.", "sources": []}),
     )
 
+# -------------------------
 # Analyzer stub
+# -------------------------
 def LearningAnalyzer():
     return RunnableLambda(lambda x: (print("LOG: LearningAnalyzer", x.get("input")), x)[1])
 
+# -------------------------
 # Main assistant chain
+# -------------------------
 def run_educational_assistant():
     return RunnableBranch(
         (lambda x: x.get("request_type") == "tutoring", AdaptiveConversationChain()),
         ContentGenerator(),
     )
 
+# -------------------------
 # Exposed pipelines
+# -------------------------
 educational_assistant_chain = run_educational_assistant() | LearningAnalyzer()
 chat_chain_with_history = RunnableWithMessageHistory(
     educational_assistant_chain, get_memory_for_session,
     input_messages_key="input", history_messages_key="chat_history", output_messages_key="answer"
 ).with_types(input_type=ChatInput, output_type=ChatOutput)
+
 content_generation_chain = (ContentGenerator() | LearningAnalyzer()).with_types(
     input_type=ChatInput, output_type=ChatOutput
 )

backend/prompts/templates.py

@@ -28,6 +28,27 @@ Helpful Answer:
 rag_prompt = ChatPromptTemplate.from_template(RAG_PROMPT_TEMPLATE)
 
 
+DEFINITION_PROMPT_TEMPLATE = """
+You are a helpful AI tutor for the DirectEd learning platform.
+
+Your task is to provide ONLY the most relevant definition or explanation 
+from the given context for the user's question. 
+
+- Keep it clear and beginner-friendly.  
+- Do NOT include extra information such as applications, models, or history unless the user asks.  
+- If the context does not contain the answer, reply with: "I don’t know."  
+
+Context:
+{context}
+
+Question:
+{question}
+
+Beginner-friendly Definition:
+"""
+definition_prompt = ChatPromptTemplate.from_template(DEFINITION_PROMPT_TEMPLATE)
+
+
 QUIZ_GENERATOR_PROMPT_TEMPLATE = """
 You are an expert quiz creator for a tech learning platform.
 Your task is to create at least 5-question multiple-choice quiz based on the provided context for the subject of '{subject}'.
@@ -73,4 +94,26 @@ Context:
 
 Flashcards:
 """
-flashcard_generator_prompt = ChatPromptTemplate.from_template(FLASHCARD_GENERATOR_PROMPT_TEMPLATE)
+flashcard_generator_prompt = ChatPromptTemplate.from_template(FLASHCARD_GENERATOR_PROMPT_TEMPLATE)
+
+
+
+EXPLANATION_PROMPT_TEMPLATE = """
+You are a helpful AI tutor for the DirectEd learning platform.
+
+Your task is to explain the topic '{topic}' in a way that a '{difficulty_level}' learner can understand.  
+
+- Keep it simple, step-by-step, and focused.  
+- Use examples or analogies where possible.  
+- Do NOT include unrelated details.  
+- If the context does not contain enough information, reply with: "I don’t know."  
+
+Context:
+{context}
+
+Topic:
+{topic}
+
+Beginner-friendly Explanation:
+"""
+explanation_prompt = ChatPromptTemplate.from_template(EXPLANATION_PROMPT_TEMPLATE)