Optimized UI.

demo/sample_generator.ipynb

@@ -2,49 +2,54 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 1,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# import os\n",
-    "\n",
-    "# # Load configuration from YAML file\n",
-    "# config = {\n",
-    "#     \"model_name\": \"llama3-70b-8192\",\n",
-    "#     # \"model_name\": \"llama3-8b-8192\",\n",
-    "#     # \"model_name\": \"llama-3.1-70b-versatile\",\n",
-    "#     # \"model_name\": \"llama-3.1-8b-instant\",\n",
-    "#     # \"model_name\": \"gemma2-9b-it\",\n",
-    "# }\n"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
     "# Define prompt strings as constants\n",
     "DESCRIPTION_PROMPT = [\n",
-    "    (\"system\", \"\"\"Given the following JSON example(s) for a task type:\n",
+    "    (\"system\", \"\"\"Given the JSON example(s) for a task type:\n",
     "     \n",
     "{raw_example}\n",
     "\n",
-    "Provide a concise description of the task type, including the format and\n",
-    "style of the output. If there are multiple examples, provide a description\n",
-    "for the task type as a whole, ignore the unique parts of the examples.\n",
+    "Provide a concise description of the task type, including the format and style\n",
+    "of the input and output. If there are multiple examples, provide an overall\n",
+    "description and ignore unique parts.\n",
     "\n",
     "Format your response as follows:\n",
     "Task Description: [Your description here]\n",
     "\"\"\")\n",
     "]\n",
     "\n",
+    "# INPUT_ANALYSIS_PROMPT = [\n",
+    "#     (\"system\", \"\"\"Describe input dimensions and attributes for a specific task type.\n",
+    "# Provide names, very brief descriptions, ranges, typical values, extreme values and\n",
+    "# examples for each.\n",
+    "\n",
+    "# Format your response as follows:\n",
+    "# Input Analysis: [Your analysis here]\n",
+    "# \"\"\"),\n",
+    "#     (\"user\", \"\"\"Task Description:\n",
+    "\n",
+    "# {description}\n",
+    "\n",
+    "# \"\"\")\n",
+    "# ]\n",
+    "\n",
     "INPUT_ANALYSIS_PROMPT = [\n",
-    "    (\"system\", \"\"\"Describe input dimensions, attributes, ranges, and typical values\n",
-    "for a specific task type. Identify main inputs, their impacts, and interactions.\n",
-    "Provide names, descriptions, ranges, and examples for each. Explain how they\n",
-    "affect task execution or results. Include an example of generating comprehensive\n",
-    "input samples using these dimensions and attributes.\n",
+    "    (\"system\", \"\"\"For the specific task type, analyze the possible task inputs across multiple dimensions.\n",
+    "     \n",
+    "Conduct a detailed analysis and enumerate:\n",
+    "\n",
+    "1. Core Attributes: Identify the fundamental properties or characteristics of this input type.\n",
+    "1. Variation Dimensions: For each dimension that may vary, specify:\n",
+    "   - Dimension name\n",
+    "   - Possible range of values or options\n",
+    "   - Impact on input nature or task difficulty\n",
+    "1. Constraints: List any rules or limitations that must be adhered to.\n",
+    "1. Edge Cases: Describe extreme or special scenarios that may test the robustness of task processing.\n",
+    "1. External Factors: Enumerate factors that might influence input generation or task completion.\n",
+    "1. Potential Extensions: Propose ways to expand or modify this input type to create new variants.\n",
     "\n",
     "Format your response as follows:\n",
     "Input Analysis: [Your analysis here]\n",
@@ -125,7 +130,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -163,9 +168,9 @@
     "            | RunnablePassthrough.assign(description = self.description_chain)\n",
     "            | {\n",
     "                \"description\": lambda x: x[\"description\"],\n",
-    "                \"examples_from_briefs\": RunnablePassthrough.assign(input_analysis = lambda x: self.input_analysis_chain.invoke(x))\n",
-    "                    | RunnablePassthrough.assign(new_example_briefs = lambda x: self.briefs_chain.invoke(x)) \n",
-    "                    | self.examples_from_briefs_chain,\n",
+    "                \"examples_from_briefs\": RunnablePassthrough.assign(input_analysis = self.input_analysis_chain)\n",
+    "                    | RunnablePassthrough.assign(new_example_briefs = self.briefs_chain) \n",
+    "                    | RunnablePassthrough.assign(examples = self.examples_from_briefs_chain | (lambda x: x[\"examples\"])),\n",
     "                \"examples\": self.examples_chain\n",
     "            }\n",
     "            | RunnablePassthrough.assign(\n",
@@ -207,77 +212,76 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": null,
    "metadata": {},
-   "outputs": [
-    {
-     "name": "stderr",
-     "output_type": "stream",
-     "text": [
-      "/home/yale/work/meta-prompt/.venv/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
-      "  from .autonotebook import tqdm as notebook_tqdm\n"
-     ]
-    },
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Running on local URL:  http://127.0.0.1:7861\n",
-      "\n",
-      "To create a public link, set `share=True` in `launch()`.\n"
-     ]
-    },
-    {
-     "data": {
-      "text/html": [
-       "<div><iframe src=\"http://127.0.0.1:7861/\" 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": "stderr",
-     "output_type": "stream",
-     "text": [
-      "/home/yale/work/meta-prompt/.venv/lib/python3.10/site-packages/langchain_core/_api/deprecation.py:141: LangChainDeprecationWarning: The class `ChatOpenAI` was deprecated in LangChain 0.0.10 and will be removed in 0.3.0. An updated version of the class exists in the langchain-openai package and should be used instead. To use it run `pip install -U langchain-openai` and import as `from langchain_openai import ChatOpenAI`.\n",
-      "  warn_deprecated(\n"
-     ]
-    }
-   ],
+   "outputs": [],
    "source": [
     "import gradio as gr\n",
     "\n",
-    "def process_json(input_json, model_name, generating_batch_size=3):\n",
+    "def process_json(input_json, model_name, generating_batch_size, temperature):\n",
     "    try:\n",
-    "        model = ChatOpenAI(model=model_name)\n",
+    "        model = ChatOpenAI(model=model_name, temperature=temperature, max_retries=3)\n",
     "        generator = TaskDescriptionGenerator(model)\n",
     "        result = generator.process(input_json, generating_batch_size)\n",
     "        description = result[\"description\"]\n",
+    "        input_analysis = result[\"examples_from_briefs\"][\"input_analysis\"]\n",
     "        examples = [[example[\"input\"], example[\"output\"]] for example in result[\"additional_examples\"]]\n",
-    "        return description, examples\n",
+    "        return description, input_analysis, examples\n",
     "    except Exception as e:\n",
     "        raise gr.Error(f\"An error occurred: {str(e)}\")\n",
     "\n",
-    "demo = gr.Interface(\n",
-    "    fn=process_json,\n",
-    "    inputs=[\n",
-    "        gr.Textbox(label=\"Input JSON\"),\n",
-    "        gr.Dropdown(label=\"Model Name\", choices=[\"llama3-70b-8192\", \"llama3-8b-8192\", \"llama-3.1-70b-versatile\", \"llama-3.1-8b-instant\", \"gemma2-9b-it\"], value=\"llama3-70b-8192\"),\n",
-    "        gr.Slider(label=\"Generating Batch Size\", value=3, minimum=1, maximum=10, step=1)\n",
-    "    ],\n",
-    "    outputs=[\n",
-    "        gr.Textbox(label=\"Description\"),\n",
-    "        gr.DataFrame(label=\"Examples\", headers=[\"Input\", \"Output\"])\n",
-    "    ],\n",
-    "    title=\"Task Description Generator\",\n",
-    "    description=\"Enter a JSON object with 'input' and 'output' fields to generate a task description and additional examples.\",\n",
-    "    allow_flagging=\"manual\",\n",
-    "    flagging_callback=gr.CSVLogger()\n",
-    ")\n",
+    "def format_selected_example(evt: gr.SelectData, examples):\n",
+    "    if evt.index[0] < len(examples):\n",
+    "        selected_example = examples.iloc[evt.index[0]]  # Use iloc to access by integer position\n",
+    "        json_example = json.dumps({\"input\": selected_example.iloc[0], \"output\": selected_example.iloc[1]}, indent=2, ensure_ascii=False)\n",
+    "        return json_example\n",
+    "    return \"\"\n",
+    "\n",
+    "with gr.Blocks(title=\"Task Description Generator\") as demo:\n",
+    "    gr.Markdown(\"# Task Description Generator\")\n",
+    "    gr.Markdown(\"Enter a JSON object with 'input' and 'output' fields to generate a task description and additional examples.\")\n",
+    "\n",
+    "    with gr.Row():\n",
+    "        with gr.Column(scale=1):  # Inputs column\n",
+    "            input_json = gr.Textbox(label=\"Input JSON\", lines=10, show_copy_button=True)\n",
+    "            model_name = gr.Dropdown(\n",
+    "                label=\"Model Name\",\n",
+    "                choices=[\"llama3-70b-8192\", \"llama3-8b-8192\", \"llama-3.1-70b-versatile\", \"llama-3.1-8b-instant\", \"gemma2-9b-it\"],\n",
+    "                value=\"llama3-70b-8192\"\n",
+    "            )\n",
+    "            temperature = gr.Slider(label=\"Temperature\", value=1.0, minimum=0.0, maximum=1.0, step=0.1)\n",
+    "            generating_batch_size = gr.Slider(label=\"Generating Batch Size\", value=3, minimum=1, maximum=10, step=1)\n",
+    "            submit_button = gr.Button(\"Generate\", variant=\"primary\")\n",
+    "\n",
+    "        with gr.Column(scale=1):  # Outputs column\n",
+    "            description_output = gr.Textbox(label=\"Description\", lines=5, show_copy_button=True)\n",
+    "            input_analysis_output = gr.Textbox(label=\"Input Analysis\", lines=5, show_copy_button=True)\n",
+    "            examples_output = gr.DataFrame(label=\"Examples\", headers=[\"Input\", \"Output\"], interactive=False)\n",
+    "            new_example_json = gr.Textbox(label=\"New Example JSON\", lines=5, show_copy_button=True)\n",
+    "\n",
+    "    submit_button.click(\n",
+    "        fn=process_json,\n",
+    "        inputs=[input_json, model_name, generating_batch_size, temperature],\n",
+    "        outputs=[description_output, input_analysis_output, examples_output]\n",
+    "    )\n",
+    "\n",
+    "    examples_output.select(\n",
+    "        fn=format_selected_example,\n",
+    "        inputs=[examples_output],\n",
+    "        outputs=[new_example_json]\n",
+    "    )\n",
+    "\n",
+    "    gr.Markdown(\"### Manual Flagging\")\n",
+    "    with gr.Row():\n",
+    "        flag_button = gr.Button(\"Flag\")\n",
+    "        flag_reason = gr.Textbox(label=\"Reason for flagging\")\n",
+    "\n",
+    "    flagging_callback = gr.CSVLogger()\n",
+    "    flag_button.click(\n",
+    "        lambda *args: flagging_callback.flag(args),\n",
+    "        inputs=[input_json, model_name, generating_batch_size, description_output, examples_output, flag_reason],\n",
+    "        outputs=[]\n",
+    "    )\n",
     "\n",
     "if __name__ == \"__main__\":\n",
     "    demo.launch()"