Multi-Agent Conversational Shopper & Dynamic Pricing Engine

from typing import TypedDict, List
from pydantic import BaseModel
from langgraph.graph import StateGraph, END
from qdrant_client import QdrantClient

class ProductItem(BaseModel):
    sku: str
    name: str
    price: float
    discounted_price: float
    stock_count: int
    confidence_score: float

class ShopperState(TypedDict):
    user_id: str
    query_text: str
    budget_max: float
    recommended_skus: List[ProductItem]
    cart_items: List[str]
    next_step: str

qdrant = QdrantClient(host="localhost", port=6333)

def intent_search_node(state: ShopperState) -> ShopperState:
    # Dense vector search in Qdrant across multimodal catalog embeddings
    search_result = qdrant.search(
        collection_name="retail_catalog",
        query_vector=[0.12, -0.45, 0.88],  # Example query embedding
        limit=5
    )
    items = []
    for res in search_result:
        items.append(ProductItem(
            sku=res.payload["sku"],
            name=res.payload["name"],
            price=res.payload["price"],
            discounted_price=res.payload["price"],
            stock_count=res.payload["stock"],
            confidence_score=res.score
        ))
    state["recommended_skus"] = items
    state["next_step"] = "dynamic_pricing_node"
    return state

def dynamic_pricing_node(state: ShopperState) -> ShopperState:
    # Apply real-time incentive discount logic
    updated_items = []
    for item in state["recommended_skus"]:
        if item.price > 100:
            item.discounted_price = round(item.price * 0.85, 2)
        updated_items.append(item)
    state["recommended_skus"] = updated_items
    return state

workflow = StateGraph(ShopperState)
workflow.add_node("intent_search", intent_search_node)
workflow.add_node("dynamic_pricing", dynamic_pricing_node)
workflow.set_entry_point("intent_search")
workflow.add_edge("intent_search", "dynamic_pricing")
workflow.add_edge("dynamic_pricing", END)