{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# app\n",
    "\n",
    "> Gradio app.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| default_exp app"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| hide\n",
    "from nbdev.showdoc import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# | export\n",
    "import copy\n",
    "import os\n",
    "import gradio as gr\n",
    "import constants\n",
    "from lv_recipe_chatbot.vegan_recipe_assistant import (\n",
    "    SYSTEM_PROMPT,\n",
    "    vegan_recipe_edamam_search,\n",
    "    VEGAN_RECIPE_SEARCH_TOOL_SCHEMA,\n",
    ")\n",
    "from openai import OpenAI, AssistantEventHandler\n",
    "from typing_extensions import override\n",
    "import json\n",
    "from functools import partial"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| hide\n",
    "import time\n",
    "from dotenv import load_dotenv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": null,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#| eval: false\n",
    "load_dotenv()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[GPT4 streaming output example on hugging face 🤗](https://huggingface.co/spaces/ysharma/ChatGPT4/blob/main/app.pyhttps://huggingface.co/spaces/ysharma/ChatGPT4/blob/main/app.py)  \n",
    "[Gradio lite let's you insert Gradio app in browser JS](https://www.gradio.app/guides/gradio-litehttps://www.gradio.app/guides/gradio-lite)  \n",
    "[Streaming output](https://www.gradio.app/main/guides/streaming-outputshttps://www.gradio.app/main/guides/streaming-outputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| eval: false\n",
    "client = OpenAI()\n",
    "assistant = client.beta.assistants.create(\n",
    "    name=\"Vegan Recipe Finder\",\n",
    "    instructions=SYSTEM_PROMPT,\n",
    "    # + \"\\nChoose the best single matching recipe to the user's query out of the vegan recipe search returned recipes\",\n",
    "    model=\"gpt-4o\",\n",
    "    tools=[VEGAN_RECIPE_SEARCH_TOOL_SCHEMA],\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class EventHandler(AssistantEventHandler):\n",
    "    @override\n",
    "    def on_event(self, event):\n",
    "        # Retrieve events that are denoted with 'requires_action'\n",
    "        # since these will have our tool_calls\n",
    "        if event.event == \"thread.run.requires_action\":\n",
    "            run_id = event.data.id  # Retrieve the run ID from the event data\n",
    "            self.handle_requires_action(event.data, run_id)\n",
    "\n",
    "    def handle_requires_action(self, data, run_id):\n",
    "        tool_outputs = []\n",
    "        for tool_call in data.required_action.submit_tool_outputs.tool_calls:\n",
    "            if tool_call.function.name == \"vegan_recipe_edamam_search\":\n",
    "                fn_args = json.loads(tool_call.function.arguments)\n",
    "                data = vegan_recipe_edamam_search(\n",
    "                    query=fn_args.get(\"query\"),\n",
    "                )\n",
    "                tool_outputs.append({\"tool_call_id\": tool_call.id, \"output\": data})\n",
    "\n",
    "        self.submit_tool_outputs(tool_outputs, run_id)\n",
    "\n",
    "    def submit_tool_outputs(self, tool_outputs, run_id):\n",
    "        client.beta.threads.runs.submit_tool_outputs_stream(\n",
    "            thread_id=self.current_run.thread_id,\n",
    "            run_id=self.current_run.id,\n",
    "            tool_outputs=tool_outputs,\n",
    "            event_handler=EventHandler(),\n",
    "        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| export\n",
    "def handle_requires_action(data):\n",
    "    tool_outputs = []\n",
    "    for tool_call in data.required_action.submit_tool_outputs.tool_calls:\n",
    "        if tool_call.function.name == \"vegan_recipe_edamam_search\":\n",
    "            fn_args = json.loads(tool_call.function.arguments)\n",
    "            data = vegan_recipe_edamam_search(\n",
    "                query=fn_args.get(\"query\"),\n",
    "            )\n",
    "            tool_outputs.append({\"tool_call_id\": tool_call.id, \"output\": data})\n",
    "    return tool_outputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def run_conversation() -> str:\n",
    "    run = client.beta.threads.runs.create_and_poll(\n",
    "        thread_id=thread.id,\n",
    "        assistant_id=assistant.id,\n",
    "    )\n",
    "    while True:\n",
    "        tool_outputs = []\n",
    "        tool_calls = (\n",
    "            []\n",
    "            if not run.required_action\n",
    "            else run.required_action.submit_tool_outputs.tool_calls\n",
    "        )\n",
    "\n",
    "        for tool_call in tool_calls:\n",
    "            if tool_call.function.name == \"vegan_recipe_edamam_search\":\n",
    "                fn_args = json.loads(tool_call.function.arguments)\n",
    "                data = vegan_recipe_edamam_search(\n",
    "                    query=fn_args.get(\"query\"),\n",
    "                )\n",
    "                tool_outputs.append({\"tool_call_id\": tool_call.id, \"output\": data})\n",
    "\n",
    "        if tool_outputs:\n",
    "            try:\n",
    "                run = client.beta.threads.runs.submit_tool_outputs_and_poll(\n",
    "                    thread_id=thread.id,\n",
    "                    run_id=run.id,\n",
    "                    tool_outputs=tool_outputs,\n",
    "                )\n",
    "                print(\"Tool outputs submitted successfully.\")\n",
    "\n",
    "            except Exception as e:\n",
    "                print(\"Failed to submit tool outputs:\", e)\n",
    "                return \"Sorry failed to run tools. Try again with a different query.\"\n",
    "\n",
    "        if run.status == \"completed\":\n",
    "            messages = client.beta.threads.messages.list(thread_id=thread.id)\n",
    "            data = messages.data\n",
    "            content = data[0].content\n",
    "            return content[0].text.value\n",
    "        time.sleep(0.05)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| export\n",
    "def run_convo_stream(thread, content: str, client: OpenAI, assistant):\n",
    "    message = client.beta.threads.messages.create(\n",
    "        thread_id=thread.id,\n",
    "        role=\"user\",\n",
    "        content=content,\n",
    "    )\n",
    "    stream = client.beta.threads.runs.create(\n",
    "        thread_id=thread.id,\n",
    "        assistant_id=assistant.id,\n",
    "        stream=True,\n",
    "    )\n",
    "    for event in stream:\n",
    "        if event.event == \"thread.message.delta\":\n",
    "            yield event.data.delta.content[0].text.value\n",
    "\n",
    "        if event.event == \"thread.run.requires_action\":\n",
    "            tool_outputs = handle_requires_action(event.data)\n",
    "            stream = client.beta.threads.runs.submit_tool_outputs(\n",
    "                run_id=event.data.id,\n",
    "                thread_id=thread.id,\n",
    "                tool_outputs=tool_outputs,\n",
    "                stream=True,\n",
    "            )\n",
    "            for event in stream:\n",
    "                if event.event == \"thread.message.delta\":\n",
    "                    yield event.data.delta.content[0].text.value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "skip\n"
     ]
    }
   ],
   "source": [
    "%%script echo  skip\n",
    "thread = client.beta.threads.create()\n",
    "\n",
    "test_msgs = [\n",
    "    \"Hello\",\n",
    "    \"What can I make with tempeh, whole wheat bread, and lettuce?\",\n",
    "]\n",
    "for m in test_msgs:\n",
    "    for txt in run_convo_stream(thread, m, client, assistant):\n",
    "        print(txt, end=\"\")\n",
    "    print()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| export\n",
    "def predict(message, history, client: OpenAI, assistant, thread):\n",
    "    # note that history is a flat list of text messages\n",
    "    reply = \"\"\n",
    "    files = message[\"files\"]\n",
    "    txt = message[\"text\"]\n",
    "\n",
    "    if files:\n",
    "        if files[-1].split(\".\")[-1] not in [\"jpg\", \"png\", \"jpeg\", \"webp\"]:\n",
    "            return \"Sorry only accept image files\"\n",
    "\n",
    "        file = message[\"files\"][-1]\n",
    "        file = client.files.create(\n",
    "            file=open(\n",
    "                file,\n",
    "                \"rb\",\n",
    "            ),\n",
    "            purpose=\"vision\",\n",
    "        )\n",
    "\n",
    "        for reply_txt in run_convo_stream(\n",
    "            thread,\n",
    "            content=[\n",
    "                {\n",
    "                    \"type\": \"text\",\n",
    "                    \"text\": \"What vegan ingredients do you see in this image? Also list out a few combinations of the ingredients that go well together. Lastly, suggest a recipe based on one of those combos using the vegan recipe seach tool.\",\n",
    "                },\n",
    "                {\"type\": \"image_file\", \"image_file\": {\"file_id\": file.id}},\n",
    "            ],\n",
    "            client=client,\n",
    "            assistant=assistant,\n",
    "        ):\n",
    "            reply += reply_txt\n",
    "            yield reply\n",
    "\n",
    "    elif txt:\n",
    "        for reply_txt in run_convo_stream(thread, txt, client, assistant):\n",
    "            reply += reply_txt\n",
    "            yield reply"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| export\n",
    "def create_demo(client: OpenAI, assistant):\n",
    "    # https://www.gradio.app/main/guides/creating-a-chatbot-fast#customizing-your-chatbot\n",
    "    # on chatbot start/ first msg after clear\n",
    "    thread = client.beta.threads.create()\n",
    "\n",
    "    # sample_images = []\n",
    "    # all_imgs = [f\"{SAMPLE_IMG_DIR}/{img}\" for img in os.listdir(SAMPLE_IMG_DIR)]\n",
    "    # for i, img in enumerate(all_imgs):\n",
    "    #     if i in [\n",
    "    #         1,\n",
    "    #         2,\n",
    "    #         3,\n",
    "    #     ]:\n",
    "    #         sample_images.append(img)\n",
    "    pred = partial(predict, client=client, assistant=assistant, thread=thread)\n",
    "    with gr.ChatInterface(\n",
    "        fn=pred,\n",
    "        multimodal=True,\n",
    "        chatbot=gr.Chatbot(\n",
    "            placeholder=\"Hello!\\nI am a animal advocate AI that is capable of recommending vegan recipes.\\nUpload an image or write a message below to get started!\"\n",
    "        ),\n",
    "    ) as demo:\n",
    "        gr.Markdown(\n",
    "            \"\"\"🔃 **Refresh the page to start from scratch**  \n",
    "        \n",
    "        Recipe search tool powered by the [Edamam API](https://www.edamam.com/)  \n",
    "        \n",
    "        ![Edamam Logo](https://www.edamam.com/assets/img/small-logo.png)\"\"\"\n",
    "        )\n",
    "\n",
    "        # clear.click(lambda: None, None, chatbot, queue=False).then(bot.reset)\n",
    "        return demo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "skip\n"
     ]
    }
   ],
   "source": [
    "%%script echo skip\n",
    "if \"demo\" in globals():\n",
    "    demo.close()\n",
    "\n",
    "demo = create_demo(client, assistant)\n",
    "demo.launch()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#| hide\n",
    "import nbdev\n",
    "\n",
    "nbdev.nbdev_export()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "local-lv-chatbot",
   "language": "python",
   "name": "local-lv-chatbot"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}