TL;DR
I deployed a semantic search system on Cloudflare's edge that costs $5-10/month instead of the typical $100-200+. It's faster, fol...
For further actions, you may consider blocking this person and/or reporting abuse
Love this, Daniel Nwaneri Proves you can build a robust RAG system without the crazy costs.
Thanks Tom! 💯
The cost difference is wild - $5 vs $200+ for similar performance.
The key was following composable MCP patterns instead of naive API wrappers. Made it both faster (365ms) and cheaper.
What stack are you using for your projects?
Thanks for the reactions @leob! Would love to hear your thoughts on the composable MCP architecture approach.
Are you building with MCP servers too?
Not yet! What do you think, should I ?
Definitely worth checking out. MCP is getting real traction right now.
Is that kind of stuff mainly programmed in Python, or is Typescript also a viable option? I'm seeing a lot of Python in this realm, and I'm not really a Python buff, never programmed in it ... but hey, if you do it "right" you don't "code" anymore anyway - you let AI generate the code for you ;-)
TypeScript works great.This entire thing is TypeScript.Python dominates the tutorials because of Jupyter notebooks and the ML heritage but for production MCP servers TypeScript is solid. Plus it works natively on Cloudflare Workers (which is why I used it).
The repo is all TS if you want to check it out: github.com/dannwaneri/vectorize-mc...
And yeah, AI-assisted coding definitely helps 😄
What are you thinking of building?
Well, nothing in particular as of yet - need to start studying and exploring, baby steps ... being able to us TS is good though!
Nice. Feel free to ping me if you have questions as you explore.
The MCP docs are pretty good, and there are some solid examples
out there now.
Good luck! 🚀
A lot of you asked how to handle the 'Exact ID' problem in this $5 stack. I just posted the solution in V2. it adds Hybrid Search and Reranking while keeping the cost at $5. Check the discussion here : dev.to/dannwaneri/my-5month-rag-sy...
Nice reality check — proves you can ship a solid production RAG for $5/month without blindly throwing money at overengineered, overpriced stacks just because everyone else does.
Thanks Willie! 💯
That's exactly the point. The "default stack" (Pinecone + OpenAI embeddings) works, but you're paying 20-40x more than you need to.
The real win: Following composable MCP patterns from Workato's framework made it faster AND cheaper. 365ms response time vs 2-4s for typical implementations.
Are you building something similar? Would love to hear what stack you're using.
A fantastic exploration and sharing experience.
More than the main idea, i love how well organized this blog is. Really impressive.
Thanks Rashid! 🙏
I spent a lot of time on the structure wanted to make sure both technical folks AND decision-makers could follow along.The key was showing real benchmarks (365ms performance) alongside the composable MCP architecture patterns.Are you working on similar edge computing projects?
Like this App!