/GENAI-company-brochure-huggingfacespaces

Modular AI tool to convert web content into markdown brochures using OpenRouter LLMs. Fully customizable and deployable via Gradio on Hugging Face Spaces.

Primary LanguagePython

title emoji colorFrom colorTo sdk sdk_version app_file pinned license
Genai Brochure Company
🐨
blue
gray
gradio
5.38.0
app.py
false
mit

Company Brochure Generator Using Gradio and Hugging Face Spaces

This tool is designed to generate a company brochure using Large Language Models (LLMs). It uses Gradio fot the user interface and Hugging Face Spaces for deployment (Public link: https://huggingface.co/spaces/Moetez1998/genai-brochure-company). The tool supports a wide range of LLMs provided by OpenRouter.

Demo

As it can be seen in the demo, the user can provide the URL of the company and select the desired LLM from the dropdown menu. The relevant information are extracted from the website using BeautifulSoup and a brochure is generated based on the extracted data.

Configuration

Config file (config.json)

Parameter Description
web_scraper_config.timeout Timeout for web scraping operations in seconds. Default is 10s
open_router_config.base_url OpenRouter API base URL
open_router_config.max_tokens Maximum number of tokens to generate in a single request. Default is 800 tokens
open_router_config.temperature Temperature for text generation sampling. Default is 0.1
open_router_config.open_router_llms Path to the OpenRouter supported LLMs
prompts.system_prompt_file Path to the system prompt file
prompts.user_prompt_file Path to the user prompt file

Supported LLMs list (open_router_llms.json)

This file defines the list of supported Large Language Models (LLMs) available through OpenRouter. Each entry includes the model's name (model_name) and its unique identifier (model_id), which is required by the OpenRouter API to specify exactly which model to use for text generation.
The following table outlines some of the currently supported models:

Model Name Organization Parameters (B) Description
DeepHermes 3 – LLaMA 3 8B Preview Nous Research 8 Fast and lightweight — Best for short content. Great performance for simple brochures.
Mistral Small 3.1 – 24B Mistral 24 Balanced — Good speed + quality tradeoff. Ideal for most brochure use cases.
Gemma 3 – 27B Google 27 Fluent and polished — Produces well-structured, professional-sounding content.
LLaMA 3.1 – 405B Instruct Meta 405 High-capacity — Best for long or dense pages. Slower, but handles complex info well.

💡 Tips: If speed matters, start with DeepHermes. For large corporate sites or detailed brochures, try LLaMA 405B. If you don't find the model for your needs, add it in the open_router_llms.json file or feel free to ask! I'm here to assist (moetez.lahdhiri@student-cs.fr).

Prompt files

Feel free to adapt the system prompt file (system_prompt.txt) for your specific needs. You can also use the user prompt file (user_prompt.txt) to customize the content generation process. Make sure to keep {company_name} and {content} placeholders in the prompt.

The Application

  1. Loads OpenRouter API key from .env file.
  2. Scrapes the webpage using beautifulsoup library.
  3. Sends content to the selected LLM for brochure generation.
  4. Displays the brochure in markdown format.
CV

Prototype Deployment on Hugging Face Spaces

You can easily deploy this prototype using Hugging Face Spaces, which provides free hosting for Gradio apps.

1. Get Your OpenRouter API Key

Visit the OpenRouter website to obtain your API Key.

  • The free plan allows up to 100 requests per day.
  • You can unlock 1,000 requests per day by adding 10 credits to your account.

💡 Tip: Use the free-tier models (i.e., model IDs ending with :free) to avoid consuming paid credits.

2. Set Up Your Hugging Face Space

Go to Hugging Face Spaces and click "Create new Space".
Configure it as follows:

  • SDK: Select Gradio as the Space SDK
  • Hardware: The default CPU Basic instance (2 vCPU, 16 GB RAM, FREE) is sufficient for this application

3. Configure Your Space

To ensure the app works correctly, make sure to:

  • ✅ Include a requirements.txt file to install necessary dependencies

  • 🔐 Add your OpenRouter API key in the "Variables and secrets" section of your Space settings

    • Key: OPENROUTER_API_KEY
    • Value: your personal API key from OpenRouter

💡 (BONUS) Other Brochure Generation Use Cases

This project is designed to be easily adaptable. Here are a few alternative use cases you can explore by tweaking the prompts or inputs:

  • 🏫 University Program Brochures
    Generate compelling overviews of academic programs from university web pages.

  • 🏥 Healthcare Services Brochures
    Summarize key services, values, and contact info from clinic or hospital websites.

  • 🎓 Bootcamps & Online Course Summaries
    Create engaging brochures for coding bootcamps, online platforms, or training centers.

  • 🌍 Non-Profit Organizations
    Generate donor-friendly content that summarizes a non-profit's mission and impact.

  • 🎤 Event or Conference Summaries
    Create brochures from event landing pages, outlining key speakers, agenda, and purpose.

✏️ To support other use cases, simply change the system and user prompts — the rest of the pipeline stays the same!