Blazor Meets Bun.js: Why I Went Hybrid for AI

A developer's take on combining the Microsoft stack with the speed of Bun for AI-powered workflows.

As someone deep in the .NET and Blazor ecosystem, I’ve always appreciated how solid and structured the Microsoft stack is. But when I started diving into AI, embeddings, and microservices, I needed something faster, lighter, and more flexible than the typical .NET approach.

That’s where Bun.js came in.

This post is a quick recap of:

Why I’m using Bun alongside Blazor
What problems I solved
How I’m using Node.js libraries like Xenova for embeddings
Why mixing the two ecosystems makes a lot of sense

🧠 The Problem: AI and LLM Workloads Are Heavy

When building features like:

Document understanding (RAG)
AI-powered chat
Voice interaction
Vector search + embeddings

…I quickly ran into limits with pure .NET. While .NET is amazing for structure and UI (Blazor Server FTW), it’s not always ideal for:

Fast background jobs
AI SDK compatibility
Running LLM tools or modern Node.js/JavaScript libraries

And honestly, I didn’t want to spend days porting working Node tools into C#. So I went hybrid.

⚡ Why Bun.js?

I chose Bun because:

It’s blazing fast – faster than Node.js for almost everything I tested
It supports Node.js libraries out of the box
It’s perfect for running lightweight microservices with zero bloat
I could use cutting-edge tools like Xenova’s all-MiniLM-L6-v2 (in JS) for embedding

The cool part? I now run Bun as a microservice backend, and Blazor/.NET calls it for everything AI-related.

🔄 My Architecture: Microservices That Fly

Here’s how it flows:

User uploads a doc in Blazor
Blazor calls a Bun.js microservice
Bun uses Xenova’s embedding model to generate vectors
The vector is stored in Qdrant
Blazor later queries Qdrant and passes results to Gemini via API

Boom — intelligent responses in real-time 🔥

🧰 Tools I’m Using

Purpose	Tool
Frontend UI	Blazor Server + Tailwind
Embeddings	Xenova’s all-MiniLM-L6-v2 via Bun
Vector Store	Qdrant + @qdrant/js-client-rest
LLM (chat/inference)	Gemini Flash Lite API
Microservice Runtime	Bun (instead of Node.js for speed)
Backend Logic	.NET 9 and Dapper

🧪 Why Mix .NET with Bun?

Simple:

Blazor is amazing for UI, security, and structure
Bun gives me raw speed + access to JS/AI tooling
Mixing both lets me move fast without losing the reliability of the .NET ecosystem

.NET doesn’t need to be your everything. Let it be your anchor, while Bun handles the async AI stuff.

🔜 What’s Next

Add LangChain.js agents to the Bun side for multi-hop logic
Add a Blazor UI for uploading, querying, and AI interaction
Package the Bun microservice setup as a reusable module for future projects

Why I’m Using Bun.js with Blazor: Speed, AI, and the Best of Both Worlds 🚀