Week 4 LangChain and LangGraph exercise solutions by Andras Nemes

4_langchain_langgraph/community_contributions/Andras_Nemes/1_lab1_exercise.ipynb

@@ -0,0 +1,207 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Lab 1 Exercise: two tools, one loop\n",
+    "\n",
+    "This notebook solves the exercise at the end of Lab 1.\n",
+    "\n",
+    "Steps:\n",
+    "\n",
+    "1. Keep the lab's `get_share_price` tool.\n",
+    "2. Add a fake `get_exchange_rate` tool.\n",
+    "3. Bind both tools to the model.\n",
+    "4. Ask for Amazon's share price in euros.\n",
+    "5. Run the tool loop by hand until the model gives a final answer.\n",
+    "\n",
+    "The answer needs a US dollar price and a USD-to-EUR rate, so both tools must run."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Imports\n",
+    "\n",
+    "Use the lab imports. `load_dotenv` finds the `.env` file in the repository root."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from dotenv import load_dotenv\n",
+    "from langchain_openai import ChatOpenAI\n",
+    "from langchain_core.messages import HumanMessage, ToolMessage\n",
+    "from langchain_core.tools import tool\n",
+    "\n",
+    "load_dotenv(override=True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Model\n",
+    "\n",
+    "Use the same model as the lab."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "llm = ChatOpenAI(model=\"gpt-5.4-mini\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Share price tool\n",
+    "\n",
+    "The lab tool returns a fake price, or `0.0` for an unknown symbol."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@tool\n",
+    "def get_share_price(symbol: str) -> float:\n",
+    "    \"\"\"Return the current share price in US dollars for a given ticker symbol.\"\"\"\n",
+    "    fake_prices = {\"AAPL\": 241.5, \"GOOG\": 168.2, \"AMZN\": 198.0}\n",
+    "    return fake_prices.get(symbol.upper(), 0.0)\n",
+    "\n",
+    "print(\"name:\", get_share_price.name)\n",
+    "print(\"args:\", get_share_price.args)\n",
+    "print(\"called directly:\", get_share_price.invoke({\"symbol\": \"AMZN\"}))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Exchange rate tool\n",
+    "\n",
+    "The new tool returns how many units of a currency equal one US dollar."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "@tool\n",
+    "def get_exchange_rate(currency: str) -> float:\n",
+    "    \"\"\"Return how many units of the given currency equal 1 US dollar.\"\"\"\n",
+    "    fake_rates = {\"EUR\": 0.92, \"GBP\": 0.79, \"JPY\": 156.0}\n",
+    "    return fake_rates.get(currency.upper(), 1.0)\n",
+    "\n",
+    "print(\"name:\", get_exchange_rate.name)\n",
+    "print(\"args:\", get_exchange_rate.args)\n",
+    "print(\"called directly:\", get_exchange_rate.invoke({\"currency\": \"EUR\"}))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Bind both tools to the model\n",
+    "\n",
+    "`bind_tools` makes both tools available to the model. The reply may contain requests in `.tool_calls` instead of a final answer."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "llm_with_tools = llm.bind_tools([get_share_price, get_exchange_rate])\n",
+    "\n",
+    "question = \"Use both tools to find Amazon's share price in euros.\"\n",
+    "response = llm_with_tools.invoke(question)\n",
+    "print(\"content:\", repr(response.content))\n",
+    "print(\"tool_calls:\", response.tool_calls)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Run the tool loop by hand\n",
+    "\n",
+    "The loop:\n",
+    "\n",
+    "1. Send the conversation to the model.\n",
+    "2. Add its reply to the conversation.\n",
+    "3. Stop if the reply has no tool calls.\n",
+    "4. Run each tool and add its result as a `ToolMessage`.\n",
+    "5. Repeat.\n",
+    "\n",
+    "The name map selects the right tool. The loop allows the model to request tools in one turn or over several turns."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "tools_by_name = {\n",
+    "    \"get_share_price\": get_share_price,\n",
+    "    \"get_exchange_rate\": get_exchange_rate,\n",
+    "}\n",
+    "\n",
+    "conversation = [HumanMessage(question)]\n",
+    "\n",
+    "while True:\n",
+    "    ai_message = llm_with_tools.invoke(conversation)\n",
+    "    conversation.append(ai_message)\n",
+    "\n",
+    "    if not ai_message.tool_calls:\n",
+    "        break\n",
+    "\n",
+    "    for call in ai_message.tool_calls:\n",
+    "        tool_to_run = tools_by_name[call[\"name\"]]\n",
+    "        result = tool_to_run.invoke(call[\"args\"])\n",
+    "        print(f\"Ran {call['name']} with {call['args']} and got {result}\")\n",
+    "        conversation.append(ToolMessage(content=str(result), tool_call_id=call[\"id\"]))\n",
+    "\n",
+    "print(\"\\nFinal answer:\")\n",
+    "print(conversation[-1].content)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "agents",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}