	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Supply Chain Optimisation Agent

Supply Chain Optimisation Agent

Author: Venkata Sudhakar

Supply chain decisions at ShopMax India involve balancing supplier pricing, delivery reliability, minimum order quantities, and lead times across dozens of vendors. An ADK supply chain agent evaluates all these factors simultaneously and recommends the optimal sourcing plan for each SKU - a task that typically takes a buyer several hours becomes a one-line query.

The agent uses four tools covering the core supply chain workflow: fetch vendor quotes, score vendor reliability, calculate economic order quantity, and generate the purchase recommendation. The model synthesises all inputs and explains its reasoning.

from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool
from google.adk.sessions import InMemorySessionService
from google.adk.runners import Runner
from google.genai.types import Content, Part

def get_vendor_quotes(sku: str) -> dict:
    """Get current vendor quotes for a ShopMax India SKU from all approved suppliers."""
    quotes = {
        "SKU-S24": [
            {"vendor": "SamsungDirect",   "unit_price_rs": 68000, "moq": 50,  "lead_days": 7},
            {"vendor": "TechDistrib",     "unit_price_rs": 66500, "moq": 100, "lead_days": 12},
            {"vendor": "MobileWholesale", "unit_price_rs": 67200, "moq": 25,  "lead_days": 5},
        ],
        "SKU-OP12": [
            {"vendor": "OnePlusOfficial", "unit_price_rs": 59000, "moq": 20,  "lead_days": 6},
            {"vendor": "TechDistrib",     "unit_price_rs": 57500, "moq": 75,  "lead_days": 14},
        ]
    }
    return {"sku": sku, "quotes": quotes.get(sku, [])}

def get_vendor_reliability_scores(vendors: list) -> dict:
    """Get historical on-time delivery and quality scores for ShopMax India vendors."""
    scores = {
        "SamsungDirect":    {"on_time_pct": 98, "defect_pct": 0.3, "rating": "A"},
        "TechDistrib":      {"on_time_pct": 82, "defect_pct": 1.2, "rating": "B"},
        "MobileWholesale":  {"on_time_pct": 91, "defect_pct": 0.8, "rating": "B+"},
        "OnePlusOfficial":  {"on_time_pct": 96, "defect_pct": 0.4, "rating": "A"},
    }
    return {v: scores.get(v, {"on_time_pct": 70, "defect_pct": 2.0, "rating": "C"})
            for v in vendors}

def calculate_order_quantity(daily_demand: float, lead_days: int,
                             holding_cost_pct: float = 0.02) -> dict:
    """Calculate recommended order quantity based on demand and lead time for ShopMax India."""
    safety_stock = daily_demand * 7  # 7-day buffer
    reorder_qty = round(daily_demand * (lead_days + 7) + safety_stock)
    return {
        "daily_demand": daily_demand,
        "safety_stock": round(safety_stock),
        "recommended_order_qty": reorder_qty
    }

def generate_purchase_recommendation(sku: str, quotes: list, reliability: dict,
                                     recommended_qty: int) -> dict:
    """Select the best vendor and generate a purchase order recommendation for ShopMax India."""
    # Score each vendor: lower price + higher reliability + meets MOQ
    best = None
    best_score = -1
    for q in quotes:
        v = q["vendor"]
        rel = reliability.get(v, {})
        if recommended_qty < q["moq"]:
            continue  # skip if we need less than MOQ
        price_score = (70000 - q["unit_price_rs"]) / 1000  # higher score for lower price
        reliability_score = rel.get("on_time_pct", 70) / 10
        lead_score = max(0, 15 - q["lead_days"])  # prefer shorter lead time
        total_score = price_score + reliability_score + lead_score
        if total_score > best_score:
            best_score = total_score
            best = q
    if not best:
        best = min(quotes, key=lambda x: x["unit_price_rs"])
    total_cost = best["unit_price_rs"] * recommended_qty
    return {
        "sku": sku,
        "recommended_vendor": best["vendor"],
        "unit_price_rs": best["unit_price_rs"],
        "order_quantity": recommended_qty,
        "total_cost_rs": total_cost,
        "lead_days": best["lead_days"],
        "reliability_rating": reliability.get(best["vendor"], {}).get("rating", "N/A")
    }

agent = LlmAgent(
    name="shopmax_supply_chain_agent",
    model="gemini-2.0-flash",
    tools=[
        FunctionTool(get_vendor_quotes),
        FunctionTool(get_vendor_reliability_scores),
        FunctionTool(calculate_order_quantity),
        FunctionTool(generate_purchase_recommendation),
    ],
    instruction=(
        "You are a ShopMax India supply chain optimisation agent. "
        "For each SKU: get vendor quotes, get reliability scores, calculate order quantity "
        "based on daily demand, then generate the purchase recommendation. "
        "Explain your vendor choice clearly."
    )
)

session_service = InMemorySessionService()
runner = Runner(agent=agent, session_service=session_service, app_name="shopmax")

for sku, demand in [("SKU-S24", 3.0), ("SKU-OP12", 2.0)]:
    session = session_service.create_session(app_name="shopmax", user_id="buyer")
    events = list(runner.run(
        user_id="buyer",
        session_id=session.id,
        new_message=Content(parts=[Part(
            text=f"Optimise sourcing for {sku} with daily demand of {demand} units."
        )])
    ))
    for event in events:
        if hasattr(event, "content") and event.content:
            for part in event.content.parts:
                if hasattr(part, "text") and part.text:
                    print(f"[{sku}] {part.text.strip()[:250]}")
    print()

It gives the following output,

[SKU-S24] Recommended vendor: MobileWholesale
Unit price: Rs 67,200 | Order qty: 84 units | Total cost: Rs 56,44,800
Lead time: 5 days | Reliability: B+ (91% on-time)

Reasoning: MobileWholesale wins on lead time (5 days vs 7 for SamsungDirect)
at only Rs 800 more per unit, and meets our MOQ of 25. TechDistrib is cheapest
but has 82% on-time delivery and 12-day lead time - too risky given current
low stock (4 days remaining).

[SKU-OP12] Recommended vendor: OnePlusOfficial
Unit price: Rs 59,000 | Order qty: 46 units | Total cost: Rs 27,14,000
Lead time: 6 days | Reliability: A (96% on-time)

Reasoning: OnePlusOfficial has better reliability (A-rated, 96% on-time) and
shorter lead time (6 vs 14 days) than TechDistrib. The Rs 1,500 premium per
unit is justified by lower defect rate and faster delivery.

The agent did not simply pick the cheapest vendor - it balanced price, lead time, reliability, and MOQ requirements. For SKU-S24 with critically low stock, it prioritised the 5-day lead time over a small price saving. For SKU-OP12, it chose the premium A-rated vendor over the cheaper but slower option. This mirrors how an experienced buyer thinks.

For ShopMax India production, connect the vendor quotes tool to the supplier portal APIs (or a vendor price feed stored in BigQuery). Store historical delivery performance in a Cloud Spanner vendor_performance table that is updated automatically when purchase orders close. Add a tool that raises the actual purchase order in the ERP system after the buyer reviews and approves the recommendation. Log all sourcing decisions with their rationale to BigQuery for quarterly supplier review meetings.

Send your comments, suggestions or queries regarding this site to [email protected].