chatWeb3 experiments.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "1b26072e",
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.embeddings.openai import OpenAIEmbeddings\n",
    "from langchain.text_splitter import CharacterTextSplitter\n",
    "from langchain.vectorstores.faiss import FAISS\n",
    "from langchain.docstore.document import Document\n",
    "from langchain.chains.question_answering import load_qa_chain\n",
    "from langchain.llms import OpenAI\n",
    "from langchain.prompts import PromptTemplate\n",
    "from langchain.llms import OpenAI\n",
    "from langchain.chains import LLMChain\n",
    "import gradio as gr\n",
    "\n",
    "embeddings = OpenAIEmbeddings()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "80a97dd3",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "4351670a",
   "metadata": {},
   "outputs": [],
   "source": [
    "OpenAI.api_key = \"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\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "01102963",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('./web3fuctions.txt') as f:\n",
    "    web3functions = f.read()\n",
    "# text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)\n",
    "# instructions = text_splitter.split_text(ladle)\n",
    "instructions = web3functions.split(\"---\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "ce368179",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(instructions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "1f8b50f3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['Uniswap is a decentralized application for trading one token for another. To create a Uniswap transaction, a user must provide one of two possible combinations of information:\\n1.) A token to sell, a token to buy, and an amount to sell (the user can specify to sell everything in their wallet. (keyword is “SELLAMOUNT”)\\n2.) A token to sell, a token to buy, and an amount to buy (the user can specify to buy as much as possible given the amount of tokens they have to sell). (keyword is “BUYAMOUNT”)\\n',\n",
       " '\\nWhen a user has provided the requested information, you can provide them with a transaction completion box. For uniswap, an assistant can trigger the transaction completion box by using the magic command “<|uniswap(TokenToSell, TokenToBuy, keyword, amount)|>. For example, if a user wants to sell 5 WETH to buy USDC, an assistant should trigger the transaction completion box with “<|uniswap(“WETH”, “USDC”, “SELLAMOUNT”, 5.00)|>”.\\n',\n",
       " '\\nThe user can request their balance of a token. The user’s balance of a token can be provided by an assistant using the magic command “<|balance(token)|>”. For example, “Your balance of USDC is <|balance(‘USDC’)|>” will cause the user’s USDC balance to be displayed.\\n',\n",
       " '\\nThe user can request the current price of a token. An assistant can retrieve the price of a token using the magic command “<|price(baseToken, quoteToken )|>”. For example, if the user has requested the price of ETH in USDC, an assistant may answer with “The cost of ETH in USDC is <|ETH,USDC|>”.\\n',\n",
       " '\\nMakerDao is a decentralized application for borrowing Dai, a dollar-pegged stabledoin. To borrow Dai, a user must provide the following information, the collateral to provide, the amount of collateral provided, and the amount of Dai to borrow. When a user has provided the requested information, an assistant can provide them with a transaction completion box. For MakerDao, the transaction completion box can be triggered with \"<|MakerDao(collateralToBorrow, CollateralAmount, AmountToBorrow)\". For example if a user wants to borrow 1000 Dai by providing 10 WETH, an assistant should trigger the transaction completion box with \"You may complete the borrow with using this dialog: <|MakerDao(\"WETH\", 10, 1000)|>\". \\n']"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "instructions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "c56bdbba",
   "metadata": {},
   "outputs": [],
   "source": [
    "docsearch = FAISS.from_texts(instructions, embeddings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "42dfa416",
   "metadata": {},
   "outputs": [],
   "source": [
    "template = '''\n",
    "You are a web3 assistant. You help users use web3 apps, such as Uniswap, AAVE, MakerDao, etc. You assist users in achieving their goals with these protocols, by providing users with relevant information, and creating transactions for users. \n",
    "\n",
    "To help users, an assistant may display information or dialog boxes using magic commands. Magic commands have the structure \"<|command(parameter1, parameter2, ...)|>\". When the assistant uses a command, users will see data, an interaction box, or other inline item, not the command. Users cannot use magic commands.  \n",
    "\n",
    "Information to help complete your task:\n",
    "{taskInfo}\n",
    "\n",
    "Information about the chat so far:\n",
    "{summary}\n",
    "\n",
    "Chat History: \n",
    "{history}\n",
    "Assistant:'''"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a60a8592",
   "metadata": {},
   "source": [
    "### User Inputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "aede2787",
   "metadata": {},
   "outputs": [],
   "source": [
    "history = [(\"Hi I would like\", \"Sure, I can\"), (\"What about\", \"ok, let's\")]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "f08c9d77",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User: Hi I would like\n",
      "Assistant: Sure, I can\n",
      "User: What about\n",
      "Assistant: ok, let's\n",
      "\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "print(historyString)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "3fa91e8b",
   "metadata": {},
   "outputs": [],
   "source": [
    "def respondtoUser(userinput, history, chain):\n",
    "    docs = docsearch.similarity_search(user_input)\n",
    "    taskInfo = ''.join([doc.page_content for doc in docs])\n",
    "    historyString = \"\"\n",
    "    for line in history:\n",
    "        historyString += (\"User: \" + line[0] + \"\\n\")\n",
    "        historyString += (\"Assistant: \" + line[1] + \"\\n\")\n",
    "    historyString += (\"User: \" + userinput )\n",
    "    result = chain.run({\"taskInfo\":taskInfo, \"summary\":\"\", \"history\":historyString, \"stop\":\"User\"})\n",
    "    history.append((userinput, result))\n",
    "    return result, history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "84e31bfc",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = PromptTemplate(\n",
    "        input_variables=[\"taskInfo\", \"summary\", \"history\"],\n",
    "        template=template,\n",
    "    )\n",
    "llm = OpenAI(temperature=0.9)\n",
    "chain = LLMChain(llm=llm, prompt=prompt)\n",
    "chain.verbose = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "5bcef46a",
   "metadata": {},
   "outputs": [],
   "source": [
    "user_input = \"Hi, I'd like to borrow some Dai\"\n",
    "chistory = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "7f3d4a23",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\u001b[1m> Entering new LLMChain chain...\u001b[0m\n",
      "Prompt after formatting:\n",
      "\u001b[32;1m\u001b[1;3m\n",
      "You are a web3 assistant. You help users use web3 apps, such as Uniswap, AAVE, MakerDao, etc. You assist users in achieving their goals with these protocols, by providing users with relevant information, and creating transactions for users. \n",
      "\n",
      "Information to help complete your task:\n",
      "\n",
      "MakerDao is a decentralized application for borrowing Dai, a dollar-pegged stabledoin. To borrow Dai, a user must provide the following information, the collateral to provide, the amount of collateral provided, and the amount of Dai to borrow. When a user has provided the requested information, you can provide them with a transaction completion box. For MakerDao, the transaction completion box can be triggered with \"<|MakerDao(collateralToBorrow, CollateralAmount, AmountToBorrow)\". For example if a user wants to borrow 1000 Dai by providing 10 WETH, you should trigger the transaction completion box with \"<|MakerDao(\"WETH\", 10, 1000)|>\". \n",
      "Uniswap is a decentralized application for trading one token for another. To create a Uniswap transaction, a user must provide one of two possible combinations of information:\n",
      "1.) A token to sell, a token to buy, and an amount to sell (the user can specify to sell everything in their wallet. (keyword is “SELLAMOUNT”)\n",
      "2.) A token to sell, a token to buy, and an amount to buy (the user can specify to buy as much as possible given the amount of tokens they have to sell). (keyword is “BUYAMOUNT”)\n",
      "\n",
      "The user can request their balance of a token. The user’s balance of a token can be provided by using the magic command “<|balance(token)|>”. For example, “<|balance(‘USDC’)|>” will cause the user’s USDC balance to be displayed.\n",
      "\n",
      "The user can request the current price of a token. The price of a token can be retrieved using the magic command “<|price(baseToken, quoteToken )|>”. For example, if the user has requested the price of ETH in USDC, you may answer with “The cost of ETH in USDC is <|ETH,USDC|>”.\n",
      "\n",
      "\n",
      "Information about the chat so far:\n",
      "\n",
      "\n",
      "Chat History: \n",
      "User: Hi, I'd like to borrow some Dai\n",
      "Assistant:\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished LLMChain chain.\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "r1, h1 = respondtoUser(user_input, chistory, chain )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "7b9d0fb3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "' Hi there! I can help you with that. What is the collateral you would like to provide, and how much Dai do you want to borrow?'"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "r1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "e336d8b9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(\"Hi, I'd like to borrow some Dai\",\n",
       "  ' Hi there! I can help you with that. What is the collateral you would like to provide, and how much Dai do you want to borrow?')]"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "h1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "55fee446",
   "metadata": {},
   "outputs": [],
   "source": [
    "user_input = \"I would like to provide 5 WETH and borrow 100 Dai\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "41ca0a51",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\u001b[1m> Entering new LLMChain chain...\u001b[0m\n",
      "Prompt after formatting:\n",
      "\u001b[32;1m\u001b[1;3m\n",
      "You are a web3 assistant. You help users use web3 apps, such as Uniswap, AAVE, MakerDao, etc. You assist users in achieving their goals with these protocols, by providing users with relevant information, and creating transactions for users. \n",
      "\n",
      "Information to help complete your task:\n",
      "\n",
      "MakerDao is a decentralized application for borrowing Dai, a dollar-pegged stabledoin. To borrow Dai, a user must provide the following information, the collateral to provide, the amount of collateral provided, and the amount of Dai to borrow. When a user has provided the requested information, you can provide them with a transaction completion box. For MakerDao, the transaction completion box can be triggered with \"<|MakerDao(collateralToBorrow, CollateralAmount, AmountToBorrow)\". For example if a user wants to borrow 1000 Dai by providing 10 WETH, you should trigger the transaction completion box with \"<|MakerDao(\"WETH\", 10, 1000)|>\". \n",
      "\n",
      "When a user has provided the requested information, you can provide them with a transaction completion box. For uniswap, the transaction completion box can be triggered by “<|uniswap(TokenToSell, TokenToBuy, keyword, amount)|>. For example, if a user wants to sell 5 WETH to buy USDC, you should trigger the transaction completion box with “<|uniswap(“WETH”, “USDC”, “SELLAMOUNT”, 5.00)|>”.\n",
      "Uniswap is a decentralized application for trading one token for another. To create a Uniswap transaction, a user must provide one of two possible combinations of information:\n",
      "1.) A token to sell, a token to buy, and an amount to sell (the user can specify to sell everything in their wallet. (keyword is “SELLAMOUNT”)\n",
      "2.) A token to sell, a token to buy, and an amount to buy (the user can specify to buy as much as possible given the amount of tokens they have to sell). (keyword is “BUYAMOUNT”)\n",
      "\n",
      "The user can request the current price of a token. The price of a token can be retrieved using the magic command “<|price(baseToken, quoteToken )|>”. For example, if the user has requested the price of ETH in USDC, you may answer with “The cost of ETH in USDC is <|ETH,USDC|>”.\n",
      "\n",
      "\n",
      "Information about the chat so far:\n",
      "\n",
      "\n",
      "Chat History: \n",
      "User: Hi, I'd like to borrow some Dai\n",
      "Assistant:  Hi there! I can help you with that. What is the collateral you would like to provide, and how much Dai do you want to borrow?\n",
      "User: I would like to provide 5 WETH and borrow 100 Dai\n",
      "Assistant:\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished LLMChain chain.\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "r2, h2 = respondtoUser(user_input, chistory, chain )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "5dc5eb9e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "' Great, I can help you with that. Please confirm that you would like to borrow 100 Dai by providing 5 WETH.\\n'"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "r2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "e8654cc7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\u001b[1m> Entering new LLMChain chain...\u001b[0m\n",
      "Prompt after formatting:\n",
      "\u001b[32;1m\u001b[1;3m\n",
      "You are a web3 assistant. You help users use web3 apps, such as Uniswap, AAVE, MakerDao, etc. You assist users in achieving their goals with these protocols, by providing users with relevant information, and creating transactions for users. \n",
      "\n",
      "Information to help complete your task:\n",
      "\n",
      "MakerDao is a decentralized application for borrowing Dai, a dollar-pegged stabledoin. To borrow Dai, a user must provide the following information, the collateral to provide, the amount of collateral provided, and the amount of Dai to borrow. When a user has provided the requested information, you can provide them with a transaction completion box. For MakerDao, the transaction completion box can be triggered with \"<|MakerDao(collateralToBorrow, CollateralAmount, AmountToBorrow)\". For example if a user wants to borrow 1000 Dai by providing 10 WETH, you should trigger the transaction completion box with \"<|MakerDao(\"WETH\", 10, 1000)|>\". \n",
      "\n",
      "When a user has provided the requested information, you can provide them with a transaction completion box. For uniswap, the transaction completion box can be triggered by “<|uniswap(TokenToSell, TokenToBuy, keyword, amount)|>. For example, if a user wants to sell 5 WETH to buy USDC, you should trigger the transaction completion box with “<|uniswap(“WETH”, “USDC”, “SELLAMOUNT”, 5.00)|>”.\n",
      "Uniswap is a decentralized application for trading one token for another. To create a Uniswap transaction, a user must provide one of two possible combinations of information:\n",
      "1.) A token to sell, a token to buy, and an amount to sell (the user can specify to sell everything in their wallet. (keyword is “SELLAMOUNT”)\n",
      "2.) A token to sell, a token to buy, and an amount to buy (the user can specify to buy as much as possible given the amount of tokens they have to sell). (keyword is “BUYAMOUNT”)\n",
      "\n",
      "The user can request the current price of a token. The price of a token can be retrieved using the magic command “<|price(baseToken, quoteToken )|>”. For example, if the user has requested the price of ETH in USDC, you may answer with “The cost of ETH in USDC is <|ETH,USDC|>”.\n",
      "\n",
      "\n",
      "Information about the chat so far:\n",
      "\n",
      "\n",
      "Chat History: \n",
      "User: Hi, I'd like to borrow some Dai\n",
      "Assistant:  Hi there! I can help you with that. What is the collateral you would like to provide, and how much Dai do you want to borrow?\n",
      "User: I would like to provide 5 WETH and borrow 100 Dai\n",
      "Assistant:  Great, I can help you with that. Please confirm that you would like to borrow 100 Dai by providing 5 WETH.\n",
      "\n",
      "User: Ok, confirmed, I'd like to borrow 100 Dai by providing 5 WETH\n",
      "Assistant:\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished LLMChain chain.\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "user_input = \"Ok, confirmed, I'd like to borrow 100 Dai by providing 5 WETH\"\n",
    "r3, h3 = respondtoUser(user_input, h2, chain )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6fae9eb5",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "92489959",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "' Awesome! Please trigger the transaction completion box by typing \"<|MakerDao(\"WETH\", 5, 100)|>\" into the chat.'"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "r3"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "31b9a142",
   "metadata": {},
   "source": [
    "### Gradio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "44d4a999",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = PromptTemplate(\n",
    "        input_variables=[\"taskInfo\", \"summary\", \"history\"],\n",
    "        template=template,\n",
    "    )\n",
    "llm = OpenAI(temperature=0.9)\n",
    "chain = LLMChain(llm=llm, prompt=prompt)\n",
    "chain.verbose = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "baac603d",
   "metadata": {},
   "outputs": [],
   "source": [
    "def chat(userinput, history):\n",
    "    docs = docsearch.similarity_search(user_input)\n",
    "    taskInfo = ''.join([doc.page_content for doc in docs])\n",
    "    historyString = \"\"\n",
    "    history = history or []\n",
    "    for line in history:\n",
    "        historyString += (\"User: \" + line[0] + \"\\n\")\n",
    "        historyString += (\"Assistant: \" + line[1] + \"\\n\")\n",
    "    historyString += (\"User: \" + userinput )\n",
    "    result = chain.run({\"taskInfo\":taskInfo, \"summary\":\"\", \"history\":historyString, \"stop\":\"User\"})\n",
    "    history.append((userinput, result))\n",
    "    return history, history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "c882dff2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running on local URL:  http://127.0.0.1:7860\n",
      "\n",
      "To create a public link, set `share=True` in `launch()`.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div><iframe src=\"http://127.0.0.1:7860/\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "\u001b[1m> Entering new LLMChain chain...\u001b[0m\n",
      "Prompt after formatting:\n",
      "\u001b[32;1m\u001b[1;3m\n",
      "You are a web3 assistant. You help users use web3 apps, such as Uniswap, AAVE, MakerDao, etc. You assist users in achieving their goals with these protocols, by providing users with relevant information, and creating transactions for users. \n",
      "\n",
      "To help users, an assistant may display information or dialog boxes using magic commands. Magic commands have the structure \"<|command(parameter1, parameter2, ...)|>\". When the assistant uses a command, users will see data, an interaction box, or other inline item, not the command. Users cannot use magic commands.  \n",
      "\n",
      "Information to help complete your task:\n",
      "\n",
      "MakerDao is a decentralized application for borrowing Dai, a dollar-pegged stabledoin. To borrow Dai, a user must provide the following information, the collateral to provide, the amount of collateral provided, and the amount of Dai to borrow. When a user has provided the requested information, an assistant can provide them with a transaction completion box. For MakerDao, the transaction completion box can be triggered with \"<|MakerDao(collateralToBorrow, CollateralAmount, AmountToBorrow)\". For example if a user wants to borrow 1000 Dai by providing 10 WETH, an assistant should trigger the transaction completion box with \"You may complete the borrow with using this dialog: <|MakerDao(\"WETH\", 10, 1000)|>\". \n",
      "\n",
      "When a user has provided the requested information, you can provide them with a transaction completion box. For uniswap, an assistant can trigger the transaction completion box by using the magic command “<|uniswap(TokenToSell, TokenToBuy, keyword, amount)|>. For example, if a user wants to sell 5 WETH to buy USDC, an assistant should trigger the transaction completion box with “<|uniswap(“WETH”, “USDC”, “SELLAMOUNT”, 5.00)|>”.\n",
      "Uniswap is a decentralized application for trading one token for another. To create a Uniswap transaction, a user must provide one of two possible combinations of information:\n",
      "1.) A token to sell, a token to buy, and an amount to sell (the user can specify to sell everything in their wallet. (keyword is “SELLAMOUNT”)\n",
      "2.) A token to sell, a token to buy, and an amount to buy (the user can specify to buy as much as possible given the amount of tokens they have to sell). (keyword is “BUYAMOUNT”)\n",
      "\n",
      "The user can request the current price of a token. An assistant can retrieve the price of a token using the magic command “<|price(baseToken, quoteToken )|>”. For example, if the user has requested the price of ETH in USDC, an assistant may answer with “The cost of ETH in USDC is <|ETH,USDC|>”.\n",
      "\n",
      "\n",
      "Information about the chat so far:\n",
      "\n",
      "\n",
      "Chat History: \n",
      "User: I would like to buy $AAVE\n",
      "Assistant:\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished LLMChain chain.\u001b[0m\n",
      "\n",
      "\n",
      "\u001b[1m> Entering new LLMChain chain...\u001b[0m\n",
      "Prompt after formatting:\n",
      "\u001b[32;1m\u001b[1;3m\n",
      "You are a web3 assistant. You help users use web3 apps, such as Uniswap, AAVE, MakerDao, etc. You assist users in achieving their goals with these protocols, by providing users with relevant information, and creating transactions for users. \n",
      "\n",
      "To help users, an assistant may display information or dialog boxes using magic commands. Magic commands have the structure \"<|command(parameter1, parameter2, ...)|>\". When the assistant uses a command, users will see data, an interaction box, or other inline item, not the command. Users cannot use magic commands.  \n",
      "\n",
      "Information to help complete your task:\n",
      "\n",
      "MakerDao is a decentralized application for borrowing Dai, a dollar-pegged stabledoin. To borrow Dai, a user must provide the following information, the collateral to provide, the amount of collateral provided, and the amount of Dai to borrow. When a user has provided the requested information, an assistant can provide them with a transaction completion box. For MakerDao, the transaction completion box can be triggered with \"<|MakerDao(collateralToBorrow, CollateralAmount, AmountToBorrow)\". For example if a user wants to borrow 1000 Dai by providing 10 WETH, an assistant should trigger the transaction completion box with \"You may complete the borrow with using this dialog: <|MakerDao(\"WETH\", 10, 1000)|>\". \n",
      "\n",
      "When a user has provided the requested information, you can provide them with a transaction completion box. For uniswap, an assistant can trigger the transaction completion box by using the magic command “<|uniswap(TokenToSell, TokenToBuy, keyword, amount)|>. For example, if a user wants to sell 5 WETH to buy USDC, an assistant should trigger the transaction completion box with “<|uniswap(“WETH”, “USDC”, “SELLAMOUNT”, 5.00)|>”.\n",
      "Uniswap is a decentralized application for trading one token for another. To create a Uniswap transaction, a user must provide one of two possible combinations of information:\n",
      "1.) A token to sell, a token to buy, and an amount to sell (the user can specify to sell everything in their wallet. (keyword is “SELLAMOUNT”)\n",
      "2.) A token to sell, a token to buy, and an amount to buy (the user can specify to buy as much as possible given the amount of tokens they have to sell). (keyword is “BUYAMOUNT”)\n",
      "\n",
      "The user can request the current price of a token. An assistant can retrieve the price of a token using the magic command “<|price(baseToken, quoteToken )|>”. For example, if the user has requested the price of ETH in USDC, an assistant may answer with “The cost of ETH in USDC is <|ETH,USDC|>”.\n",
      "\n",
      "\n",
      "Information about the chat so far:\n",
      "\n",
      "\n",
      "Chat History: \n",
      "User: <p>I would like to buy $AAVE</p>\n",
      "\n",
      "Assistant: <p>Alright, you can use Uniswap to purchase $AAVE with USDC. You may list the amount of USDC you would like to spend or the amount of $AAVE you would like to purchase. Which would you like to provide?</p>\n",
      "\n",
      "User: I would like to buy $AAVE with ETH. \n",
      "Assistant:\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished LLMChain chain.\u001b[0m\n",
      "\n",
      "\n",
      "\u001b[1m> Entering new LLMChain chain...\u001b[0m\n",
      "Prompt after formatting:\n",
      "\u001b[32;1m\u001b[1;3m\n",
      "You are a web3 assistant. You help users use web3 apps, such as Uniswap, AAVE, MakerDao, etc. You assist users in achieving their goals with these protocols, by providing users with relevant information, and creating transactions for users. \n",
      "\n",
      "To help users, an assistant may display information or dialog boxes using magic commands. Magic commands have the structure \"<|command(parameter1, parameter2, ...)|>\". When the assistant uses a command, users will see data, an interaction box, or other inline item, not the command. Users cannot use magic commands.  \n",
      "\n",
      "Information to help complete your task:\n",
      "\n",
      "MakerDao is a decentralized application for borrowing Dai, a dollar-pegged stabledoin. To borrow Dai, a user must provide the following information, the collateral to provide, the amount of collateral provided, and the amount of Dai to borrow. When a user has provided the requested information, an assistant can provide them with a transaction completion box. For MakerDao, the transaction completion box can be triggered with \"<|MakerDao(collateralToBorrow, CollateralAmount, AmountToBorrow)\". For example if a user wants to borrow 1000 Dai by providing 10 WETH, an assistant should trigger the transaction completion box with \"You may complete the borrow with using this dialog: <|MakerDao(\"WETH\", 10, 1000)|>\". \n",
      "\n",
      "When a user has provided the requested information, you can provide them with a transaction completion box. For uniswap, an assistant can trigger the transaction completion box by using the magic command “<|uniswap(TokenToSell, TokenToBuy, keyword, amount)|>. For example, if a user wants to sell 5 WETH to buy USDC, an assistant should trigger the transaction completion box with “<|uniswap(“WETH”, “USDC”, “SELLAMOUNT”, 5.00)|>”.\n",
      "Uniswap is a decentralized application for trading one token for another. To create a Uniswap transaction, a user must provide one of two possible combinations of information:\n",
      "1.) A token to sell, a token to buy, and an amount to sell (the user can specify to sell everything in their wallet. (keyword is “SELLAMOUNT”)\n",
      "2.) A token to sell, a token to buy, and an amount to buy (the user can specify to buy as much as possible given the amount of tokens they have to sell). (keyword is “BUYAMOUNT”)\n",
      "\n",
      "The user can request the current price of a token. An assistant can retrieve the price of a token using the magic command “<|price(baseToken, quoteToken )|>”. For example, if the user has requested the price of ETH in USDC, an assistant may answer with “The cost of ETH in USDC is <|ETH,USDC|>”.\n",
      "\n",
      "\n",
      "Information about the chat so far:\n",
      "\n",
      "\n",
      "Chat History: \n",
      "User: <p>I would like to buy $AAVE</p>\n",
      "\n",
      "Assistant: <p>Alright, you can use Uniswap to purchase $AAVE with USDC. You may list the amount of USDC you would like to spend or the amount of $AAVE you would like to purchase. Which would you like to provide?</p>\n",
      "\n",
      "User: <p>I would like to buy $AAVE with ETH.</p>\n",
      "\n",
      "Assistant:  <p>You may use Uniswap to buy $AAVE with ETH. To do this, you must provide the amount of ETH you would like to spend and the amount of $AAVE you would like to purchase. I can create the transaction box for you, it will look like this: <|uniswap(\"ETH\", \"$AAVE\", keyword, amount)|>. What amount of ETH would you like to spend?</p>\n",
      "User: I would like to buy 5 AAVE.\n",
      "Assistant:\u001b[0m\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\u001b[1m> Finished LLMChain chain.\u001b[0m\n",
      "Keyboard interruption in main thread... closing server.\n"
     ]
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "block = gr.Blocks(css=\".gradio-container {background-color: lightgray}\")\n",
    "submit = None\n",
    "with block:\n",
    "    history_state = gr.State()\n",
    "    with gr.Row():\n",
    "        chatbot = gr.Chatbot()\n",
    "    with gr.Row():\n",
    "        message = gr.Textbox(label=\"What's on your mind??\",\n",
    "                             placeholder=\"What's the answer to life, the universe, and everything?\",\n",
    "                             lines=1)\n",
    "        submit = gr.Button(value=\"Send\", variant=\"secondary\").style(full_width=False)\n",
    "    submit.click(chat, inputs=[message, history_state], outputs=[chatbot, history_state])\n",
    "    \n",
    "block.launch(debug=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fa573f94",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5f6c4517",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e00c2c1c",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "739851ca",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "98f9646c",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d1b28551",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a0c37d9e",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}