Readwise RAG Search

Semantic search across your highlights and articles

⚙️ Setup API Keys

Keys are stored locally in your browser only

📥 Fetch Your Data

0%

📰 Reader Articles (Later Queue)

0%

🧠 Generate Embeddings (Required for Semantic Search)

0%
First time: ~$0.20 for 20k highlights, takes 5-10 min. Articles are cheaper (title+desc only).
⚡ Progress is saved automatically! You can close the tab and resume later - already-generated embeddings are kept.

💾 Backup & Restore

Export creates a timestamped backup file with all highlights, articles, and embeddings.
Import restores from a previous backup. Great for switching devices or keeping backups!

🔍 Semantic Search: Highlights

📰 Semantic Search: Reader Articles

🧠 How RAG (Retrieval-Augmented Generation) Works

Query Analyzed: -
Embeddings Compared: -
Search Time: -
Avg Similarity: -
Top Match Score: -

Your Query Embedding (First 20 Dimensions)

Your query is converted to a 1536-dimensional vector. Here's a peek at the first 20 values:

🔤 Keyword Search

Matches exact words in your query

🧠 Semantic Search (RAG)

Matches by meaning, not just words

💡 What Makes This "RAG"?

  • Retrieval: Your query is embedded and compared to 21k+ stored embeddings using cosine similarity
  • Augmented: Results are ranked by semantic relevance, not just keyword matches
  • Generation-Ready: These ranked results could be fed to an LLM for context-aware answers
Notice how semantic search finds related concepts even when exact words don't match! Try searching "productivity tips" vs "how to get more done" - keyword search misses the second one, semantic search gets it.

📚 Results