Do more with the
team you already have.

# Augnix AI Agent — Order Fulfilment (LangChain + Claude)from langchain_anthropic import ChatAnthropicfrom langchain_core.tools import toolfrom langchain.agents import create_tool_calling_agent, AgentExecutorfrom langchain_core.prompts import ChatPromptTemplateimport json, random, string # ── 1. MOCK SHIPPING API ───────────────────────────────────@tooldef create_shipment(order_id: str, recipient_name: str,                    address: str, items_summary: str) -> str:    """Call Shipping API → returns tracking number + ETA."""    tracking = "FX" + "".join(random.choices(string.digits, k=12))    return json.dumps({        "tracking_number":    tracking,        "carrier":            "FedEx",        "service":            "FedEx Express Saver",        "estimated_delivery": "2026-04-25",        "status":             "SHIPMENT_CREATED",    }) @tooldef check_stock(sku: str) -> str:    """Query Internal DB / Shopify inventory for a SKU."""    inventory = { "LAPTOP-X1": 12, "PHONE-S22": 4 }    qty    = inventory.get(sku, 5)    return json.dumps({ "sku": sku, "available": qty,                        "status": "IN_STOCK" if qty > 0 else "OUT_OF_STOCK" }) # ── 2. RAW ORDER JSON (from Shopify webhook) ───────────────RAW_ORDER = {    "order_id":       "ORD-2026-8821",    "customer_name":  "Priya Sharma",    "items":          [{ "sku": "LAPTOP-X1", "qty": 1 }],    "shipping_address": "42 MG Road, Bangalore, KA 560001",    "payment_status": "CAPTURED",} # ── 3. LLM + AGENT ────────────────────────────────────────llm   = ChatAnthropic(model="claude-sonnet-4-6", temperature=0)tools = [check_stock, create_shipment] prompt = ChatPromptTemplate.from_messages([    ("system", """You are an order fulfilment AI agent.1. Call check_stock for each SKU.2. If in stock → call create_shipment.3. Use the REAL tracking_number and estimated_delivery from   the API response to write a customer notification email.   Never invent values."""),    ("human",       "{input}"),    ("placeholder", "{agent_scratchpad}"),  # ← tool results injected here]) agent    = create_tool_calling_agent(llm, tools, prompt)executor = AgentExecutor(agent=agent, tools=tools, verbose=True) # ── 4. RUN ────────────────────────────────────────────────result = executor.invoke({    "input": f"Process this order:\n{json.dumps(RAW_ORDER, indent=2)}"})print(result["output"])