src/rag_graph.py

# src/rag_graph.py
from langgraph.graph import StateGraph, END
from langgraph.checkpoint.memory import MemorySaver
from langchain_core.prompts import ChatPromptTemplate

from src.core.graph_state import GraphState
from src.core.embeddings import load_embeddings
from src.core.llm import load_llm
from src.vector_store.vector_store import build_vector_store
from src.config.config import K_OFFSET, MMR_LAMBDA
from src.exceptions import VectorStoreNotInitializedError, LLMInvocationError


class ProjectRAGGraph:
    def __init__(self):
        self.embeddings = load_embeddings()
        self.llm = load_llm()
        self.vector_store = None
        self.pdf_count = 0
        self.memory = MemorySaver()
        self.workflow = self._build_graph()

    def process_documents(self, pdf_paths, original_names=None):
        self.pdf_count = len(pdf_paths)
        self.vector_store = build_vector_store(
            pdf_paths,
            self.embeddings,
            original_names
        )

    # ---------- Graph Nodes ----------

    def retrieve(self, state: GraphState):
        if not self.vector_store:
            raise VectorStoreNotInitializedError("Vector store not initialized")

        k_value = max(1, self.pdf_count + K_OFFSET)

        retriever = self.vector_store.as_retriever(
            search_type="mmr",
            search_kwargs={"k": k_value, "lambda_mult": MMR_LAMBDA}
        )

        documents = retriever.invoke(state["question"])
        return {"context": documents}

    def generate(self, state: GraphState):
        try:
            prompt = ChatPromptTemplate.from_template(
                """
                You are an expert Project Analyst.
                Answer ONLY using the provided context.
                If the answer is not present, say "I don't know".

                Context:
                {context}

                Question:
                {question}
                """
            )

            formatted_context = "\n\n".join(
                doc.page_content for doc in state["context"]
            )

            chain = prompt | self.llm
            response = chain.invoke({
                "context": formatted_context,
                "question": state["question"]
            })

            return {"answer": response.content}

        except Exception as e:
            raise LLMInvocationError(f"LLM failed: {e}")

    # ---------- Graph Build ----------

    def _build_graph(self):
        workflow = StateGraph(GraphState)
        workflow.add_node("retrieve", self.retrieve)
        workflow.add_node("generate", self.generate)
        workflow.set_entry_point("retrieve")
        workflow.add_edge("retrieve", "generate")
        workflow.add_edge("generate", END)
        return workflow.compile(checkpointer=self.memory)

    def query(self, question: str, thread_id: str):
        config = {"configurable": {"thread_id": thread_id}}
        result = self.workflow.invoke({"question": question}, config=config)
        return result["answer"]