SQL-RAG-Demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "03a5c00c",
   "metadata": {},
   "source": [
    "## In this Notebook, we will create a basic Q&A chain and Agent over a SQL database"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d06633c3",
   "metadata": {},
   "source": [
    "### Architecture\n",
    "\n",
    "At a high-level, the steps of any SQL chain and agent are:\n",
    "\n",
    "- `Convert question to SQL query using the LLM Model.`\n",
    "- `Execute the SQL query.`\n",
    "- `getting the Model to respond to user input using the query results.`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "5ddeeaf6",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.utilities import SQLDatabase\n",
    "from langchain.chat_models import ChatOpenAI\n",
    "from langchain.chains import create_sql_query_chain\n",
    "from langchain_openai import ChatOpenAI\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "203dcbe2",
   "metadata": {},
   "outputs": [],
   "source": [
    "llm = ChatOpenAI(temperature=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "a5ac620a",
   "metadata": {},
   "outputs": [],
   "source": [
    "host = '0.0.0.0'\n",
    "port = '5432'\n",
    "username = 'odin'\n",
    "password = 'odin'\n",
    "db_name = 'analytics'\n",
    "postgres_connection = f\"postgresql://{username}:{password}@{host}:{port}/{db_name}\"\n",
    "\n",
    "db = SQLDatabase.from_uri(postgres_connection, sample_rows_in_table_info=2)\n",
    "chain = create_sql_query_chain(llm, db)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "796c438c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "postgresql\n",
      "['customer', 'sale']\n"
     ]
    }
   ],
   "source": [
    "print(db.dialect)\n",
    "print(db.get_usable_table_names())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "8fb4c912",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'[(1000,)]'"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "db.run(\"SELECT count(*) FROM customer LIMIT 10;\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "fb6515e1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'SELECT COUNT(customerid) AS total_customers FROM customer;'"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "response = chain.invoke({\"question\": \"How many customer are there\"})\n",
    "response"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "1d66e095",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'[(1000,)]'"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "db.run(response)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "b3d18b40",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SELECT COUNT(*) \n",
      "FROM customer \n",
      "WHERE city = 'Bronx' \n",
      "LIMIT 5;\n"
     ]
    }
   ],
   "source": [
    "response = chain.invoke({\"question\": \"How many customers are from Bronx\"})\n",
    "print(response)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "eb84b1ba",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(2,)]\n"
     ]
    }
   ],
   "source": [
    "print(db.run(response))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "13a5754c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SELECT \"state_\", COUNT(*) as state_count\n",
      "FROM customer\n",
      "GROUP BY \"state_\"\n",
      "ORDER BY state_count DESC\n",
      "LIMIT 1;\n"
     ]
    }
   ],
   "source": [
    "response = chain.invoke({\"question\": \"Which state are customers most from?\"})\n",
    "print(response)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "393e48fd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[('CA', 79)]\n"
     ]
    }
   ],
   "source": [
    "print(db.run(response))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "id": "c293f9d4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SELECT c.customerid, c.firstname, c.lastname, s.shipping\n",
      "FROM customer c\n",
      "JOIN sale s ON c.customerid = s.customerid\n",
      "ORDER BY s.shipping DESC\n",
      "LIMIT 5;\n"
     ]
    }
   ],
   "source": [
    "response = chain.invoke({\"question\": \"Give me the top 5 customers who paid the highest shipping charges\"})\n",
    "print(response)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "id": "a1fc6919",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Payment:\n",
    "    id: int\n",
    "    name: str\n",
    "    surname: str\n",
    "    shipping: float\n",
    "\n",
    "    def __init__(\n",
    "            self,\n",
    "            id: int,\n",
    "            name: str,\n",
    "            surname: str,\n",
    "            shipping: float) -> None:\n",
    "        self.id = id\n",
    "        self.name = name\n",
    "        self.surname = surname\n",
    "        self.shipping = shipping\n",
    "\n",
    "    @staticmethod\n",
    "    def from_tuple(data: tuple):\n",
    "        return Payment(\n",
    "            id = data[0],\n",
    "            name = data[1],\n",
    "            surname = data[2],\n",
    "            shipping = data[3],\n",
    "        )\n",
    "    \n",
    "    def __str__(self) -> str:\n",
    "        return f\"\"\"\n",
    "            id: {self.id}\n",
    "            name: {self.name}\n",
    "            surname: {self.surname}\n",
    "            shipping: {self.shipping}\n",
    "        \"\"\"\n",
    "    \n",
    "    def to_dict(self) -> dict:\n",
    "        return {\n",
    "            'id': self.id,\n",
    "            'name': self.name,\n",
    "            'surname': self.surname,\n",
    "            'shipping': self.shipping\n",
    "        }\n",
    "\n",
    "class Customer:\n",
    "    customerid: int \n",
    "    firstname: str \n",
    "    lastname: str \n",
    "    streetaddress: str \n",
    "    city: str \n",
    "    state_: str \n",
    "    postalcode: str \n",
    "    country: str\n",
    "    phone: str\n",
    "\n",
    "    def __init__(\n",
    "            self,\n",
    "            customerid: int,\n",
    "            firstname: str ,\n",
    "            lastname: str ,\n",
    "            streetaddress: str, \n",
    "            city: str ,\n",
    "            state_: str, \n",
    "            postalcode: str, \n",
    "            country: str,\n",
    "            phone: str  \n",
    "        ) -> None:\n",
    "        self.customerid = customerid\n",
    "        self.firstname = firstname\n",
    "        self.lastname = lastname\n",
    "        self.streetaddress = streetaddress \n",
    "        self.city =  city\n",
    "        self.state_ = state_\n",
    "        self.postalcode = postalcode\n",
    "        self.country = country\n",
    "        self.phone = phone\n",
    "\n",
    "    @staticmethod\n",
    "    def from_tuple(customer: tuple) :\n",
    "        return Customer(\n",
    "                customerid = customer[0],\n",
    "                firstname = customer[1],\n",
    "                lastname = customer[2],\n",
    "                streetaddress = customer[3], \n",
    "                city =  customer[4],\n",
    "                state_ = customer[5],\n",
    "                postalcode = customer[6],\n",
    "                country = customer[7],\n",
    "                phone = customer[8]\n",
    "            )\n",
    "\n",
    "    def to_dict(self) -> dict:\n",
    "        return {\n",
    "           \"customerid\": self.customerid,\n",
    "           \"firstname\": self.firstname,\n",
    "           \"lastname\": self.lastname,\n",
    "           \"streetaddress\": self.streetaddress,\n",
    "           \"city\": self.city,\n",
    "           \"state_\": self.state_,\n",
    "           \"postalcode\": self.postalcode,\n",
    "           \"country\": self.country,\n",
    "           \"phone\": self.phone\n",
    "        }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "id": "f301db69",
   "metadata": {},
   "outputs": [],
   "source": [
    "res = db.run(response)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "id": "e9941ff6",
   "metadata": {},
   "outputs": [],
   "source": [
    "import ast\n",
    "\n",
    "payments = []\n",
    "\n",
    "for item in ast.literal_eval(res):\n",
    "    payments.append(Payment.from_tuple(data=item))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "id": "4fe6d1da",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "     id         name     surname  shipping\n",
      "0  1096       Meliha     Handzic     59.37\n",
      "1  1950         Jack      Freund     54.45\n",
      "2   724  Epaminondas  Kapetanios     54.41\n",
      "3  3979         Glen    Quintana     54.39\n",
      "4  1385        Peter       Pille     52.88\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "users_dicts = [item.to_dict() for item in payments]\n",
    "\n",
    "df = pd.DataFrame(users_dicts)\n",
    "\n",
    "print(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "id": "6cae0326",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt \n",
    "\n",
    "# Extracting 'name' and 'value' columns\n",
    "names = df['name']\n",
    "values = df['shipping']\n",
    "\n",
    "# Plotting the data\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(names, values, color='skyblue')\n",
    "plt.xlabel('Names')\n",
    "plt.ylabel('Shippings')\n",
    "plt.title('Values by Name')\n",
    "plt.xticks(rotation=45)\n",
    "plt.tight_layout()\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "id": "38513acd",
   "metadata": {},
   "outputs": [],
   "source": [
    "response = chain.invoke({\"question\": \"Give me the last 10 customer ordered by first name\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "id": "ffd1ed79",
   "metadata": {},
   "outputs": [],
   "source": [
    "res = db.run(response)\n",
    "\n",
    "customers = []\n",
    "\n",
    "for item in ast.literal_eval(res):\n",
    "    customers.append(Customer.from_tuple(item))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "id": "5bcc0d17",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>customerid</th>\n",
       "      <th>firstname</th>\n",
       "      <th>lastname</th>\n",
       "      <th>streetaddress</th>\n",
       "      <th>city</th>\n",
       "      <th>state_</th>\n",
       "      <th>postalcode</th>\n",
       "      <th>country</th>\n",
       "      <th>phone</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>727</td>\n",
       "      <td>Zhiyan</td>\n",
       "      <td>Hu</td>\n",
       "      <td>Po Box 3726</td>\n",
       "      <td>El Centro</td>\n",
       "      <td>CA</td>\n",
       "      <td>92244</td>\n",
       "      <td>USA</td>\n",
       "      <td>(760) 352-4913</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>500</td>\n",
       "      <td>Zaheeruddin</td>\n",
       "      <td>Asif</td>\n",
       "      <td>46 Mary Ave</td>\n",
       "      <td>East Providence</td>\n",
       "      <td>RI</td>\n",
       "      <td>02914</td>\n",
       "      <td>USA</td>\n",
       "      <td>(401) 434-1321</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2006</td>\n",
       "      <td>Yvonne</td>\n",
       "      <td>Weber</td>\n",
       "      <td>2416 S 13th St</td>\n",
       "      <td>Temple</td>\n",
       "      <td>TX</td>\n",
       "      <td>76504</td>\n",
       "      <td>USA</td>\n",
       "      <td>(254) 770-0397</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1129</td>\n",
       "      <td>Yungchang</td>\n",
       "      <td>Ku</td>\n",
       "      <td>1951 Liko Pl</td>\n",
       "      <td>Wailuku</td>\n",
       "      <td>HI</td>\n",
       "      <td>96793</td>\n",
       "      <td>USA</td>\n",
       "      <td>(808) 244-5817</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1276</td>\n",
       "      <td>Yuan-Hong</td>\n",
       "      <td>Liu</td>\n",
       "      <td>3 Twiller St</td>\n",
       "      <td>Albany</td>\n",
       "      <td>NY</td>\n",
       "      <td>12209</td>\n",
       "      <td>USA</td>\n",
       "      <td>(518) 462-4608</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>3906</td>\n",
       "      <td>Young-Woo</td>\n",
       "      <td>Sohn</td>\n",
       "      <td>805 E 8th Ave</td>\n",
       "      <td>Kingman</td>\n",
       "      <td>KS</td>\n",
       "      <td>67068</td>\n",
       "      <td>USA</td>\n",
       "      <td>(620) 532-5217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>927</td>\n",
       "      <td>Yoichi</td>\n",
       "      <td>Terashita</td>\n",
       "      <td>2110 9th St</td>\n",
       "      <td>Sanger</td>\n",
       "      <td>CA</td>\n",
       "      <td>93657</td>\n",
       "      <td>USA</td>\n",
       "      <td>(559) 875-3682</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>819</td>\n",
       "      <td>Yishu.Aileen</td>\n",
       "      <td>Ding</td>\n",
       "      <td>851 Lott Rd</td>\n",
       "      <td>Pocatello</td>\n",
       "      <td>ID</td>\n",
       "      <td>83201</td>\n",
       "      <td>USA</td>\n",
       "      <td>(208) 637-8209</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1801</td>\n",
       "      <td>Yaniv</td>\n",
       "      <td>Vakrat</td>\n",
       "      <td>33 Booker St</td>\n",
       "      <td>Hampton</td>\n",
       "      <td>VA</td>\n",
       "      <td>23663</td>\n",
       "      <td>USA</td>\n",
       "      <td>(757) 788-8500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>2372</td>\n",
       "      <td>Yang</td>\n",
       "      <td>Lee</td>\n",
       "      <td>3501 S Randi Pl</td>\n",
       "      <td>Tucson</td>\n",
       "      <td>AZ</td>\n",
       "      <td>85730</td>\n",
       "      <td>USA</td>\n",
       "      <td>(520) 885-2081</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customerid     firstname   lastname    streetaddress             city  \\\n",
       "0         727        Zhiyan         Hu      Po Box 3726        El Centro   \n",
       "1         500   Zaheeruddin       Asif      46 Mary Ave  East Providence   \n",
       "2        2006        Yvonne      Weber   2416 S 13th St           Temple   \n",
       "3        1129     Yungchang         Ku     1951 Liko Pl          Wailuku   \n",
       "4        1276     Yuan-Hong        Liu     3 Twiller St           Albany   \n",
       "5        3906     Young-Woo       Sohn    805 E 8th Ave          Kingman   \n",
       "6         927        Yoichi  Terashita      2110 9th St           Sanger   \n",
       "7         819  Yishu.Aileen       Ding      851 Lott Rd        Pocatello   \n",
       "8        1801         Yaniv     Vakrat     33 Booker St          Hampton   \n",
       "9        2372          Yang        Lee  3501 S Randi Pl           Tucson   \n",
       "\n",
       "  state_ postalcode country           phone  \n",
       "0     CA      92244     USA  (760) 352-4913  \n",
       "1     RI      02914     USA  (401) 434-1321  \n",
       "2     TX      76504     USA  (254) 770-0397  \n",
       "3     HI      96793     USA  (808) 244-5817  \n",
       "4     NY      12209     USA  (518) 462-4608  \n",
       "5     KS      67068     USA  (620) 532-5217  \n",
       "6     CA      93657     USA  (559) 875-3682  \n",
       "7     ID      83201     USA  (208) 637-8209  \n",
       "8     VA      23663     USA  (757) 788-8500  \n",
       "9     AZ      85730     USA  (520) 885-2081  "
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "customers_dicts = [item.to_dict() for item in customers]\n",
    "\n",
    "pd.DataFrame(customers_dicts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "id": "f3182878",
   "metadata": {},
   "outputs": [],
   "source": [
    "response = chain.invoke({\"question\": \"Give me firts 3 sales shipping and the customer name\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "id": "bf174fb9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SELECT s.shipping, c.firstname, c.lastname\n",
      "FROM sale s\n",
      "JOIN customer c ON s.customerid = c.customerid\n",
      "ORDER BY s.saleid\n",
      "LIMIT 3;\n"
     ]
    }
   ],
   "source": [
    "print(response)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "id": "1c9b398a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(44.37, 'Xiafeng', 'Wang'), (9.95, 'Herschel', 'Webster'), (39.05, 'John', 'Hebeler')]\n"
     ]
    }
   ],
   "source": [
    "res = db.run(response)\n",
    "\n",
    "print(res)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "id": "97481c99",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "You are a PostgreSQL expert. Given an input question, first create a syntactically correct PostgreSQL query to run, then look at the results of the query and return the answer to the input question.\n",
      "Unless the user specifies in the question a specific number of examples to obtain, query for at most 5 results using the LIMIT clause as per PostgreSQL. You can order the results to return the most informative data in the database.\n",
      "Never query for all columns from a table. You must query only the columns that are needed to answer the question. Wrap each column name in double quotes (\") to denote them as delimited identifiers.\n",
      "Pay attention to use only the column names you can see in the tables below. Be careful to not query for columns that do not exist. Also, pay attention to which column is in which table.\n",
      "Pay attention to use CURRENT_DATE function to get the current date, if the question involves \"today\".\n",
      "\n",
      "Use the following format:\n",
      "\n",
      "Question: Question here\n",
      "SQLQuery: SQL Query to run\n",
      "SQLResult: Result of the SQLQuery\n",
      "Answer: Final answer here\n",
      "\n",
      "Only use the following tables:\n",
      "\n",
      "CREATE TABLE customer (\n",
      "\tcustomerid BIGINT, \n",
      "\tfirstname VARCHAR(1024), \n",
      "\tlastname VARCHAR(1024), \n",
      "\tstreetaddress VARCHAR(1024), \n",
      "\tcity VARCHAR(1024), \n",
      "\tstate_ VARCHAR(1024), \n",
      "\tpostalcode VARCHAR(1024), \n",
      "\tcountry VARCHAR(1024), \n",
      "\tphone VARCHAR(1024)\n",
      ")\n",
      "\n",
      "/*\n",
      "2 rows from customer table:\n",
      "customerid\tfirstname\tlastname\tstreetaddress\tcity\tstate_\tpostalcode\tcountry\tphone\n",
      "12\tJames\tDominick\t4 Mast Ct\tPortsmouth\tVA\t23703\n"
     ]
    }
   ],
   "source": [
    "prompt_with_context = chain.get_prompts()[0].partial(table_info=context[\"table_info\"])\n",
    "print(prompt_with_context.pretty_repr()[:1500])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "id": "33c14cc5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['table_info', 'table_names']\n",
      "\n",
      "CREATE TABLE customer (\n",
      "\tcustomerid BIGINT, \n",
      "\tfirstname VARCHAR(1024), \n",
      "\tlastname VARCHAR(1024), \n",
      "\tstreetaddress VARCHAR(1024), \n",
      "\tcity VARCHAR(1024), \n",
      "\tstate_ VARCHAR(1024), \n",
      "\tpostalcode VARCHAR(1024), \n",
      "\tcountry VARCHAR(1024), \n",
      "\tphone VARCHAR(1024)\n",
      ")\n",
      "\n",
      "/*\n",
      "2 rows from customer table:\n",
      "customerid\tfirstname\tlastname\tstreetaddress\tcity\tstate_\tpostalcode\tcountry\tphone\n",
      "12\tJames\tDominick\t4 Mast Ct\tPortsmouth\tVA\t23703\tUSA\t(757) 686-2992\n",
      "17\tFredric\tWagner\t11 May St\tEnfield\tNH\t03748\tUSA\t(603) 632-7463\n",
      "*/\n",
      "\n",
      "\n",
      "CREATE TABLE sale (\n",
      "\tsaleid BIGINT, \n",
      "\tsaledate VARCHAR(1024), \n",
      "\tcustomerid BIGINT, \n",
      "\ttax DOUBLE PRECISION, \n",
      "\tshipping DOUBLE PRECISION\n",
      ")\n",
      "\n",
      "/*\n",
      "2 rows from sale table:\n",
      "saleid\tsaledate\tcustomerid\ttax\tshipping\n",
      "191228\t2013-12-31\t1134\t0.0\t44.37\n",
      "191229\t2013-12-31\t14958\t0.0\t9.95\n",
      "*/\n"
     ]
    }
   ],
   "source": [
    "context = db.get_context()\n",
    "print(list(context))\n",
    "print(context[\"table_info\"])"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}