ai-brochure-generator

AI-Powered Web Content Synthesis and Autonomous Brokerage This repository contains a modular system designed for autonomous web navigation, content extraction, and professional summary synthesis. The architecture is engineered to decouple data acquisition from cognitive processing, ensuring high maintainability and the capacity to transition between various inference backends with minimal structural modification.

Architectural Overview The application adheres to a strictly modular design pattern to maintain a clear separation of concerns:

Data Acquisition Layer (scraper.py): Responsible for low-level HTTP requests and HTML parsing. It utilizes BeautifulSoup4 to sanitize the Document Object Model (DOM) by decomposing non-informative elements, such as scripts and styles, to maximize the information density provided to the language model.

Cognitive Inference Layer (agent.py): Orchestrates communication with the Large Language Model (LLM). It utilizes the OpenAI-compatible API provided by Ollama to execute Llama 3.2 locally, ensuring data privacy and eliminating external API latency.

Orchestration Layer (main.py): Serves as the primary entry point, managing the execution pipeline and facilitating data transfer between the scraper and the inference engine.

Technical Specifications Directory Structure Plaintext ai-brochure-generator/ ├── src/ │ ├── scraper.py # Logic for web crawling and DOM sanitization │ ├── agent.py # LLM prompt engineering and API orchestration │ └── main.py # Application entry point and orchestrator ├── requirements.txt # Verified project dependencies └── README.md # Technical documentation Dependency Management The system requires the following Python libraries:

openai: Standardized communication protocol for the LLM.

beautifulsoup4: Robust HTML parsing and data extraction.

requests: Management of network protocols and content retrieval.

Implementation Guide Prerequisites A functioning installation of Ollama is required. The specific model weights for Llama 3.2 must be localized prior to runtime:

Bash ollama pull llama3.2 Installation and Execution Clone the repository:

Bash git clone https://github.com/[USERNAME]/ai-brochure-generator.git cd ai-brochure-generator Initialize the environment:

Bash python3 -m pip install -r requirements.txt Execute the orchestrator:

Bash python3 -m src.main Methodology The agent employs a deterministic approach to content summarization. Upon receiving a target URL, the system retrieves the raw HTML, extracts the primary body text, and performs a truncation at 2,000 characters. This methodology ensures that the most critical business information is prioritized for the LLM's reasoning engine while respecting the computational constraints of local hardware (Apple Silicon).

Research and Development Roadmap Integration of Playwright for asynchronous JavaScript rendering and Single Page Application (SPA) support.

Implementation of a recursive link-selection algorithm to allow autonomous navigation through deep site hierarchies.

Support for structured export formats including Markdown and LaTeX-generated PDFs.