NLWeb aims to simplify the creation of conversational interfaces for websites by leveraging existing structured data formats like Schema.org. It targets website owners and developers seeking to enable natural language interactions for both human users and AI agents, thereby establishing a foundational layer for an "AI Web."

How It Works

NLWeb operates on two core components: a simple protocol for natural language interaction and a data format utilizing JSON and Schema.org for responses. It abstracts website content, particularly lists of items, into a queryable format. The Model Context Protocol (MCP) is central, enabling AI assistants to interact with websites via a core ask method, with responses structured using Schema.org. This approach is analogous to HTTP for HTML, providing a standardized communication layer.

Quick Start & Requirements

• Install/Run: The README does not provide a specific installation command but details the components available in the repository: core service, LLM/vector database connectors, data ingestion tools, and a web server frontend with a basic UI.
• Prerequisites: Tested with Windows, macOS, Linux. Supports various LLMs (OAI, Deepseek, Gemini, Anthropic) and vector stores (Qdrant, Snowflake, Milvus, Azure AI Search).
• Documentation:
  • Getting Started: Hello world on your laptop
  • Core Concepts: Life of a Chat Query
  • REST Interface: REST interface

Highlighted Details

• Leverages Schema.org, RSS, and other semi-structured formats used by over 100 million websites.
• Designed to be platform-agnostic regarding operating systems, LLMs, and vector stores.
• Intended to be lightweight and scalable, deployable from laptops to cloud clusters.
• Encourages integration with live databases to avoid data freshness issues.

Maintenance & Community

• Project is managed by Microsoft.
• Contribution guidance is available.
• Contact for questions: NLWeb Support.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive MIT license allows for commercial use and integration into closed-source applications.

Limitations & Caveats

The repository currently lacks continuous integration and automated deployments. Cloud deployment guides for GCP and AWS are listed as "coming soon."