TMSA Spotlight

This application generates social media posts (Facebook, LinkedIn, X/Twitter, and Instagram) to feature Transportation Marketing and Sales Association (TMSA) member companies. It uses:

• Requests + BeautifulSoup for extracting homepage text,
• Ollama for local LLM inference (Llama 3.1), and
• Gradio for a user-friendly interface allowing up to 12 company/website entries at once.

When you run this app, each company gets a text file with four social media posts (one for each platform) in the output folder, according to TMSA’s branding and style guidelines.

---

Table of Contents

1. Prerequisites
2. Installation
3. Usage
4. How It Works
5. Troubleshooting
6. License

---

Prerequisites

1. Python 3.11+.
2. Ollama installed and configured on your Windows system, including the Llama 3.1 model (llama3.1:latest or your preferred variant).
   • You can verify installation in the Windows terminal:

     where ollama

   • If ollama is not on your PATH, add it or refer to Ollama’s documentation.
3. Basic Python environment with the following libraries installed:
   • requests
   • beautifulsoup4
   • gradio

---

Installation

1. Clone or download this repository.

2. In your command prompt (Windows) or terminal, navigate to the project directory and install the Python dependencies:

   pip install requests beautifulsoup4 gradio ollama

3. Check if Ollama is properly installed and accessible:

   where ollama

   If not found, follow Ollama’s documentation to install and set up on Windows.

4. (Optional) Create and activate a virtual environment to keep dependencies isolated:

   python -m venv venv
   venv\Scripts\activate
   pip install -r requirements.txt

---

Usage

1. Run the Application From your terminal, run:

   python app.py

   This launches a local Gradio interface (usually at http://127.0.0.1:7860).

2. Enter Company/Website Pairs

   • In the Gradio app, enter up to 12 company names and their respective website URLs.
   • Click Generate Posts.

3. Review Output

   • For each valid (Company Name, Website URL) pair, the app will:
     1. Extract homepage text (using requests + BeautifulSoup).
     2. Generate four social media posts (Facebook, LinkedIn, X, Instagram) by calling Ollama sequentially.
     3. Save all four posts in a single text file named {CompanyName_YYYYMMDD}.txt inside the output directory.
   • Progress and log messages appear in the Gradio textbox and in the app.log file.

---

How It Works

1. Text Extraction:

   • The code uses requests.get() to fetch the homepage HTML and BeautifulSoup to parse it, extracting text from <p> and <h1>–<h6> tags.
   • If any error occurs (e.g., invalid URL, timeout), the process logs the error and skips that company.

2. Generating Social Media Posts:

   • The script calls ollama.chat() four times in sequence (Facebook, LinkedIn, X/Twitter, Instagram).
   • Each post uses unique prompts tailored for that social platform, combined with TMSA brand guidelines and a snippet of extracted homepage text.
   • The app waits for each LLM response before moving on to the next.

3. Saving Output:

   • The final text file for each company includes all four posts, labeled with the social network.
   • By default, the files are placed in an output directory, which is created automatically if it doesn’t exist.

---

Troubleshooting

• Ollama not found:

  • Run where ollama on Windows. If it doesn’t show a path, ensure you’ve installed Ollama and added it to your PATH.
  • Confirm you have the correct Llama model: llama3.1:latest. If different, update the model_name variable in the code.

• Python library not installed:

  • Run pip show requests (or pip list | findstr requests on Windows).
  • If missing, run pip install requests. Repeat for beautifulsoup4, gradio, etc.

• Time out / Connection errors:

  • Some websites or restrictive network environments can lead to timeouts. Increase timeout in requests.get(url, timeout=10) or try again later.

• Large Webpages / Long Summaries:

  • The code currently truncates homepage text to the first 1500 characters. Adjust this in shortened_homepage_context = homepage_text[:1500] if needed.

• Output not appearing:

  • Check your output folder.
  • Check app.log for any error messages.

---

License

This project is provided via the MIT license. It is intended as a demonstration for creating a local AI agent application with Ollama, requests, BeautifulSoup, and Gradio.

Feel free to adapt or expand this code for your own internal or commercial use, abiding by any relevant licensing.