A full-stack AI-powered tool that fetches and summarizes web articles using LangChain.js and OpenAI GPT-3.5.

This project is an experimental AI pipeline for summarizing articles, built to explore LangChain’s capabilities in handling long text, contextual summarization, and structured outputs.

• Fetch article content dynamically from any valid URL.
• Chunk-based contextual summarization for long articles, processed progressively.
• Extract Key Points and generate Suggested Titles based on user-selected options.
• Dynamic prompt engineering with structured JSON outputs.
• Frontend built with Next.js, allowing users to input URLs, select summary options, and view results in real-time.
• Support for fake vs real pipeline, for easy testing without API costs.

1. The user submits a URL and selects what they want (summary, key points, title).
2. The article content is fetched and split into manageable chunks.
3. Each chunk is summarized contextually, updating a global summary progressively (manual memory approach).
4. Once all chunks are processed, the final summary is analyzed again to extract key points and suggest a title.
5. The result is returned as a structured JSON, allowing for clean UI presentation.

• Contextual Summarization: Summarizing progressively for better coherence.
• Structured Output: Returning JSON makes it easier to work with and extend.
• API Experimentation: This setup allows for learning how to manage AI-driven processes practically.

• Node.js / Next.js – Full-stack JavaScript.
• LangChain.js – Orchestration for LLMs.
• OpenAI GPT-3.5 – Language model for summarization.
• Cheerio.js – For extracting text from HTML content.
• CSS Modules – Scoped styling for clean UI.

This project is in progress and will be expanded to explore other AI techniques in the future.

Planned enhancements:

• Multi-document support.
• File uploads (PDF, DOCX).
• Rate limiting and caching.
• GitHub Actions for automated testing.

👉 Project Board