Skip to main content

Overview

Semantic Search (v0.7) lets you find previously converted content using natural language instead of exact keywords. Ask for “articles about database optimization” and Web2MD will surface relevant results — even if those exact words never appeared in the content.
Semantic Search is a Pro-only feature.

How it works

1

Convert pages as usual

Every time you convert a page, Web2MD automatically generates a vector embedding of the content. No extra steps required.
2

Search with natural language

Open History and type a query in plain English (or any language). Web2MD matches your query against all stored embeddings.
3

Get ranked results

Results are ranked by semantic relevance, not keyword frequency. The most conceptually similar conversions appear first.

Cross-language matching

Semantic Search works across languages. An English query will match Chinese, Japanese, Spanish, or any other language content — and vice versa.
Query languageContent languageWorks?
EnglishEnglishYes
EnglishChineseYes
ChineseEnglishYes
AnyAnyYes
This is especially useful if you read content in multiple languages. Search once, find everything relevant regardless of language.
After converting a page, Web2MD automatically shows up to 3 related articles from your history. This surfaces connections you might not have noticed — a blog post from last month that covers the same topic, or a documentation page with overlapping concepts. Related content appears below the Preview Card after each conversion.

Technical details

Embeddings

Powered by OpenAI embeddings, generated automatically on every conversion. Stored securely in your account.

Vector search

Uses pgvector for fast similarity search across your entire conversion history.

Where to use it

Open the History tab in the Web2MD popup and type your query in the search bar. Results update as you type.

Zero configuration

Embeddings are generated automatically — you don’t need to enable anything or provide API keys. As long as you have a Pro plan, every conversion is searchable from the moment it’s saved.
Embedding generation happens server-side and typically completes within a few seconds after conversion.