Anyone who's built a Retrieval-Augmented Generation (RAG) pipeline knows the pain: stitching together data stores, vector databases, embedding models, LLMs, and then keeping it all updated... it's a lot.
AutoRAG aims to completely abstract that complexity away. It's a fully managed RAG pipeline built right on Cloudflare's developer platform.
Here’s the basic idea:
You point AutoRAG at your data source (starting with Cloudflare R2 buckets).
Cloudflare automatically handles the entire backend: ingesting files, converting them (even images!), chunking, creating embeddings (via Workers AI), storing them (in Vectorize), and crucially, continuously re-indexing as your data updates.
You query it via API or a native Worker binding to get AI-generated answers grounded in your specific data.
Key aspects:
⚙️ Fully Managed: Cloudflare takes care of the complex infra and maintenance. 🔄 Automated Indexing: Keeps your RAG context up-to-date effortlessly. ☁️ Built on Cloudflare: Seamlessly integrates with R2, Vectorize, Workers AI, and AI Gateway. 🖱️ Simple Setup: Designed to get started with just a few clicks in the dashboard. 🆓 Free During Open Beta: Available on all Cloudflare plans, including the free tier!
It is a super streamlined way to add powerful, context-aware AI features to apps, especially if you're already using Cloudflare.
AutoRAG sounds like a huge win for teams wanting RAG without the setup headache. How customizable is the retrieval logic or ranking if you need domain-specific tuning?
Replies
Hi everyone!
Anyone who's built a Retrieval-Augmented Generation (RAG) pipeline knows the pain: stitching together data stores, vector databases, embedding models, LLMs, and then keeping it all updated... it's a lot.
AutoRAG aims to completely abstract that complexity away. It's a fully managed RAG pipeline built right on Cloudflare's developer platform.
Here’s the basic idea:
You point AutoRAG at your data source (starting with Cloudflare R2 buckets).
Cloudflare automatically handles the entire backend: ingesting files, converting them (even images!), chunking, creating embeddings (via Workers AI), storing them (in Vectorize), and crucially, continuously re-indexing as your data updates.
You query it via API or a native Worker binding to get AI-generated answers grounded in your specific data.
Key aspects:
⚙️ Fully Managed: Cloudflare takes care of the complex infra and maintenance.
🔄 Automated Indexing: Keeps your RAG context up-to-date effortlessly.
☁️ Built on Cloudflare: Seamlessly integrates with R2, Vectorize, Workers AI, and AI Gateway.
🖱️ Simple Setup: Designed to get started with just a few clicks in the dashboard.
🆓 Free During Open Beta: Available on all Cloudflare plans, including the free tier!
It is a super streamlined way to add powerful, context-aware AI features to apps, especially if you're already using Cloudflare.
Hi all, thank you for checking out AutoRAG!
I'm a part of the AutoRAG team, feel free to let me know if you have any questions or feedback. :)
Also it'd be so cool to see any projects you've built with AutoRAG. Please share if you have any!
Platus (YC F24)
AutoRAG sounds like a huge win for teams wanting RAG without the setup headache. How customizable is the retrieval logic or ranking if you need domain-specific tuning?
@felixgerlach we have chunking, retrieval, and system prompt settings. You can see a full list here: https://developers.cloudflare.com/autorag/configuration/. We are planning to add reranking soon.
Do you have any other customization settings you'd like to see?
Just take my upvote
Loving how seamless this sounds! ⚙️ AutoRAG takes the heavy lifting out of RAG—just connect your data and let Cloudflare handle the rest.
Loving how seemless it looks. Just tried it. Felt awesome