بسم الله الرحمن الرحيم
In the name of Allah, the Most Gracious, the Most Merciful

What is Cloudflare AutoRAG?

In short, it’s an integrated system designed to save developers time and effort when building Retrieval-Augmented Generation (RAG) applications by automating most of the steps and enabling direct querying of data via an API or Workers.

test

---

What Steps Are Automated?

Data ingestion
Converting documents of various formats to Markdown
Chunking the resulting text
Embedding the chunks into vector representations
Creating a vector database and storing the embeddings
Enabling similarity search and queries over the stored data via API or Workers
Improving queries and retrieval results using Large Language Models (LLMs)

All the developer needs to do is upload the required documents or data to the R2 storage and wait for the indexing process to complete — then they can immediately start querying.
The service currently offers a free tier for testing.

What Are the Current Limitations?

PDF Documents in Arabic
When testing, I noticed that converting Arabic PDF documents to Markdown doesn’t work correctly, leading to failed storage and incorrect querying.
Suggested solution: Manually convert Arabic documents to Markdown before uploading them to R2.
Metadata Search Not Supported
There’s no built-in way to search using metadata.
Workaround: Organize documents into folders and use folder names or groups for filtering and querying.
Document Size Limit (4MB)
The supported document size is relatively small.
Solution: Split large documents into smaller parts before uploading to R2.