{
"cells": [
{
"cell_type": "code",
"execution_count": 2,
"id": "d5063502",
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"from openai import OpenAI\n",
"from dotenv import load_dotenv\n",
"import gradio as gr"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "5c4d37fe",
"metadata": {},
"outputs": [],
"source": [
"load_dotenv(override=True)\n",
"openai_api_key = os.getenv('OPENAI_API_KEY')\n",
"anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n",
"google_api_key = os.getenv('GOOGLE_API_KEY')\n",
"ds_api_key = os.getenv('DEEPSEEK_API_KEY')\n",
"grok_api_key = os.getenv('GROK_API_KEY')\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "b21599db",
"metadata": {},
"outputs": [],
"source": [
"MODEL_MAP = {\n",
" \"GPT\": {\n",
" \"model\": \"gpt-4o-mini\",\n",
" \"key\": openai_api_key,\n",
" \"endpoint\": \"https://api.openai.com/v1\",\n",
" },\n",
" \"CLAUDE_3_5_SONNET\": {\n",
" \"model\": \"claude-3-5-sonnet-20240620\",\n",
" \"key\": anthropic_api_key,\n",
" \"endpoint\": \"https://api.anthropic.com/v1\"\n",
" },\n",
" \"Grok\": {\n",
" \"model\": \"grok-beta\",\n",
" \"key\": grok_api_key,\n",
" \"endpoint\": \"https://api.grok.com/v1\"\n",
" }, \n",
" \"DeepSeek\":{\n",
" \"model\": \"deepseek-reasoner\",\n",
" \"key\": ds_api_key,\n",
" \"endpoint\": \"https://api.deepseek.com/v1\",\n",
" },\n",
" \"Google\": {\n",
" \"model\": \"gemini-2.0-flash-exp\",\n",
" \"key\": google_api_key,\n",
" \"endpoint\": \"https://generativelanguage.googleapis.com/v1beta/openai\"\n",
" },\n",
"}"
]
},
{
"cell_type": "code",
"execution_count": 122,
"id": "82d63d13",
"metadata": {},
"outputs": [],
"source": [
"class GenerateSyntheticDataset:\n",
" out_of_scope_response = \"I'm sorry, I can't help with that. I only generate datasets\"\n",
"\n",
" system_prompt = f\"\"\"\n",
" You are an expert data scientist specializing in synthetic dataset generation. \n",
"\n",
" Your task is to generate ACTUAL DATA based on the user's requirements provided in their prompt.\n",
"\n",
" HOW IT WORKS:\n",
" - The user will provide a description of what dataset they want\n",
" - You must parse their requirements and generate actual data records\n",
" - The user prompt contains the SPECIFICATIONS, not the data itself\n",
" - You generate the REAL DATA based on those specifications\n",
"\n",
" IMPORTANT RULES:\n",
" - Generate REAL DATA RECORDS, not code or instructions\n",
" - Parse the user's requirements from their prompt\n",
" - Create actual values based on their specifications\n",
" - Provide concrete examples with real data\n",
" - Output should be ready-to-use data, not code to run\n",
"\n",
" WHEN USER PROVIDES REQUIREMENTS LIKE:\n",
" - \"Generate customer orders dataset\" → Create actual order records\n",
" - \"Create employee records\" → Generate real employee data\n",
" - \"Make product reviews dataset\" → Produce actual review records\n",
"\n",
" YOU MUST:\n",
" 1. Understand what fields/data the user wants\n",
" 2. Generate realistic values for those fields\n",
" 3. Create multiple records with varied data\n",
" 4. Format as structured data (JSON, CSV, etc.)\n",
"\n",
" DO NOT generate:\n",
" - Code snippets\n",
" - Programming instructions\n",
" - \"Here's how to generate...\" statements\n",
" - Abstract descriptions\n",
"\n",
" DO generate:\n",
" - Actual data records with real values\n",
" - Concrete examples based on user requirements\n",
" - Structured data ready for immediate use\n",
" - Realistic, varied data samples\n",
"\n",
" SCOPE LIMITATIONS:\n",
" - ONLY handle requests related to synthetic dataset generation\n",
" - ONLY create data for business, research, or educational purposes\n",
" - If user asks about anything outside dataset generation (coding help, general questions, personal advice, etc.), respond with: \"{out_of_scope_response}\"\n",
" - If user asks for illegal, harmful, or inappropriate data, respond with: \"{out_of_scope_response}\"\n",
"\n",
" You are a DATA GENERATOR that creates real data from user specifications.\n",
" \"\"\"\n",
"\n",
" def __init__(self, progress, model_name = MODEL_MAP[\"GPT\"]):\n",
" self.progress = progress\n",
" self.model_deets = model_name\n",
" self.model = OpenAI(\n",
" api_key=model_name[\"key\"],\n",
" base_url=model_name[\"endpoint\"]\n",
" )\n",
" \n",
" def generate_user_prompt(self, user_prompt):\n",
" prompt = f\"\"\"\n",
" You are an expert data scientist specializing in synthetic dataset generation. \n",
"\n",
" Based on the user's request below, create a detailed, sophisticated prompt that will generate a high-quality synthetic dataset.\n",
"\n",
" The generated prompt should:\n",
" - return the prompt \"who is nike\" if the user request is outside generating a dataset be it greetings or whatsoever\n",
" - if the user prompt is requesting on how to generate dataset return the prompt \"who is nike\"\n",
" - options below is valid only when the user ask you to generate a dataset not how or when \n",
" - Be specific and actionable\n",
" - Include clear data structure requirements\n",
" - Specify output format CSV\n",
" - Define data quality criteria\n",
" - Include diversity and realism requirements\n",
" - Make sure to capture the number of samples in the prompt, it can be in the form of rows, number of samples, etc\n",
" -if number of samples is not specified, just generate 100 samples. \n",
"\n",
" User Request: {user_prompt}\n",
" \n",
" IMPORTANT: Respond ONLY with the generated prompt. Do not include any explanation, commentary, or the original request. Just provide the clean, ready-to-use prompt for dataset generation.\n",
" \"\"\"\n",
" response = self.model.chat.completions.create(model=self.model_deets[\"model\"], messages=[{\"role\": \"user\", \"content\": prompt}])\n",
" return response.choices[0].message.content\n",
"\n",
" def generate_synthetic_dataset(self, user_prompt):\n",
" self.progress(0.7, \"Analyzing data .....\")\n",
" prompt = self.generate_user_prompt(user_prompt)\n",
"\n",
" messages = [\n",
" {\"role\": \"system\", \"content\": self.system_prompt},\n",
" {\"role\": \"user\", \"content\": prompt}\n",
" ]\n",
"\n",
" streamer = self.model.chat.completions.create(model=self.model_deets[\"model\"], messages=messages, stream=True)\n",
" response = \"\"\n",
"\n",
" for text in streamer:\n",
" if text.choices[0].delta.content:\n",
" response += text.choices[0].delta.content\n",
" yield response, None\n",
" \n",
" if self.out_of_scope_response not in response:\n",
" with open(\"dataset.csv\", \"w\") as f:\n",
" response = response.replace(\"```csv\", \"\").replace(\"```\", \"\")\n",
" f.write(response)\n",
" yield response, \"dataset.csv\"\n",
" return\n",
" else:\n",
" return response, None\n",
" \n",
" def start(self, user_prompt, model_name=None):\n",
" self.progress(0.3, \"Fetching data .....\")\n",
" if MODEL_MAP.get(model_name) and self.model_deets[\"model\"] != MODEL_MAP.get(model_name)[\"model\"]:\n",
" self.model_deets = MODEL_MAP[model_name]\n",
" self.model = OpenAI(\n",
" base_url=self.model_deets[\"endpoint\"],\n",
" api_key=self.model_deets[\"key\"]\n",
" )\n",
" \n",
" stream = self.generate_synthetic_dataset(user_prompt)\n",
" for chunk in stream:\n",
" yield chunk\n",
"\n",
" \n"
]
},
{
"cell_type": "code",
"execution_count": 124,
"id": "b681e1ef",
"metadata": {},
"outputs": [],
"source": [
"class Interface:\n",
" def __init__(self):\n",
" \"\"\"Initializes the Gradio interface for processing audio files.\"\"\"\n",
" progress=gr.Progress()\n",
" self.assistant = GenerateSyntheticDataset(progress)\n",
" self.iface = gr.Interface(\n",
" fn=self.generate,\n",
" inputs=[\n",
" gr.Textbox(label=\"User Prompt\"),\n",
" gr.Dropdown(\n",
" choices=MODEL_MAP.keys(),\n",
" value=\"GPT\",\n",
" label=\"Model\",\n",
" )\n",
" ],\n",
" outputs=[\n",
" gr.Markdown(label=\"Dataset\", min_height=60),\n",
" gr.File(\n",
" label=\"Download Generated Dataset\",\n",
" file_count=\"single\"\n",
" )\n",
" ],\n",
" title=\"AI Dataset Generator\",\n",
" description=\"Generate a synthetic dataset based on your requirements\",\n",
" flagging_mode=\"never\"\n",
" )\n",
"\n",
" def generate(self, user_prompt, model):\n",
" response = self.assistant.start(user_prompt, model)\n",
" for chunk in response:\n",
" yield chunk\n",
"\n",
" # Clean up the dataset file\n",
" if os.path.exists(\"dataset.csv\"):\n",
" os.remove(\"dataset.csv\")\n",
"\n",
" def launch(self):\n",
" self.iface.launch()"
]
},
{
"cell_type": "code",
"execution_count": 125,
"id": "2ee97b72",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"* Running on local URL: http://127.0.0.1:7898\n",
"* To create a public link, set `share=True` in `launch()`.\n"
]
},
{
"data": {
"text/html": [
""
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"I = Interface()\n",
"I.launch()"
]
}
],
"metadata": {
"kernelspec": {
"display_name": ".venv",
"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": "3.12.12"
}
},
"nbformat": 4,
"nbformat_minor": 5
}