DEV Community: Atul Tripathi

Why building RAG apps feels harder than it should and what I built to fix it

Atul Tripathi — Mon, 26 Jan 2026 12:15:29 +0000

Why building RAG apps feels harder than it should (and what I built to fix it)

Atul Tripathi ・ Jan 26

#webdev #rag #ai #buildinpublic

Why building RAG apps feels harder than it should (and what I built to fix it)

Atul Tripathi — Mon, 26 Jan 2026 12:13:50 +0000

I kept seeing the same pattern whenever someone attempted to create an AI app 🤖

The idea was always exciting 🚀
The demo was always ambitious 💡

Then everything came to a halt ⛔

This happened not because the idea was flawed, but due to messy foundations 🧱
PDF ingestion broke down 📄❌
Scrapers failed on random websites 🕷️⚠️
Embeddings were slow 🐌
The UI wasn’t set up correctly 🧩❌
Every small change brought about three new bugs 🐛🐛🐛

People didn’t quit because they were lazy 😴
They quit because the tools weren’t made for builders who need speed ⚡

So, I created Fastrag 🛠️

Fastrag is a ready-made backbone for RAG-based AI products 🧠
You can plug in PDFs or URLs, and it turns them into a searchable knowledge base with a working chat interface instantly 🔍💬

There’s no need to struggle with infrastructure 🏗️
No more stitching together half-working scripts 🧵❌
You won’t spend your first week on plumbing 🚿

You can start building features on day one 📅
Fastrag takes care of the rest 🤝

Because the real bottleneck in AI isn’t ideas 💭
It’s execution speed 🏎️

Build your AI product this weekend without burning out. 🔥

Uploaded a 120-page PDF instant AI chat. If you want the same setup, FastRAG is live - https://www.fastrag.live/

Atul Tripathi — Sat, 03 Jan 2026 05:37:48 +0000

FastRAG | Ship your AI App in days 🚀

Stop building RAG from scratch. Grab the starter kit with Pinecone & LangChain pre-configured.

fastrag.live

Uploaded a 120-page PDF instant AI chat. If you want the same setup, FastRAG is live - fastrag.live

Atul Tripathi — Sat, 03 Jan 2026 05:37:04 +0000

Uploaded a 120-page PDF instant AI chat. If you want the same setup, FastRAG is live - https://www.fastrag.live/

Atul Tripathi — Sat, 03 Jan 2026 04:59:51 +0000

FastRAG | Ship your AI App in days 🚀

Stop building RAG from scratch. Grab the starter kit with Pinecone & LangChain pre-configured.

fastrag.live

How to Build a "Chat with Website" App using Next.js, LangChain, and Cheerio

Atul Tripathi — Mon, 15 Dec 2025 09:42:10 +0000

Atul Tripathi

Dec 15 '25

How to Build a "Chat with Website" App using Next.js, LangChain, and Cheerio 🦜🔗

#nextjs #tutorial #ai #javascript

Comments

2 min read

How to Build a "Chat with Website" App using Next.js, LangChain, and Cheerio 🦜🔗

Atul Tripathi — Mon, 15 Dec 2025 09:41:34 +0000

Building RAG (Retrieval Augmented Generation) apps usually starts with PDFs. 📄
But let's be honest: users really want to chat with live URLs—documentation, wikis, and blogs. 🌐

I spent this weekend adding a Web Scraper to my RAG Starter Kit. Here is the technical breakdown of how I built it, so you can do it too. 👇

🛑 The Problem with Scraping for LLMs

You can't just fetch(url) and pass the HTML to GPT-4.

Too much noise: Navbars, footers, and ads waste tokens. 💸
Context Window: Raw HTML is huge and confuses the model.
Headless Browsers: Tools like Puppeteer are heavy and often timeout on serverless functions (like Vercel). ⏳

🛠 The Stack

Framework: Next.js 14
Scraper: Cheerio (via LangChain). It parses HTML like jQuery, making it lightweight and fast. ⚡️
Vector DB: Pinecone (Serverless).

Step 1: The Scraper Logic 🕷️

We use CheerioWebBaseLoader from LangChain. It grabs the raw HTML and lets us select only the body or specific content tags (like <article>), ignoring the junk.

import { CheerioWebBaseLoader } from "langchain/document_loaders/web/cheerio";

export async function scrapeUrl(url) {
  // 1. Load the URL
  const loader = new CheerioWebBaseLoader(url, {
    selector: "p, h1, h2, h3, article", // 🎯 Only grab text content
  });

  const docs = await loader.load();
  return docs;
}

Step 2: The Cleaning (Smart Chunking) 🧹

LLMs need manageable chunks of text. If you cut a sentence in half, you lose context. We use RecursiveCharacterTextSplitter.


import { RecursiveCharacterTextSplitter } from "langchain/text_splitter";

const splitter = new RecursiveCharacterTextSplitter({
  chunkSize: 1000, // 📏 Tokens per chunk
  chunkOverlap: 200, // 🔗 Overlap to preserve context across chunks
});

const splitDocs = await splitter.splitDocuments(docs);

Step 3: The Cost Hack (1024 Dimensions) 💡

This is the most important part! 💰

By default, OpenAI's embedding models output 1536 dimensions. But Pinecone charges based on storage size.

OpenAI's new text-embedding-3-small allows you to "shorten" the dimensions with minimal accuracy loss.

I configured my implementation to force 1024 dimensions:


const embeddings = new OpenAIEmbeddings({
  modelName: "text-embedding-3-small",
  dimensions: 1024, // 📉 Saves ~33% on storage costs
});

✅ The Result

We now have a clean pipeline: URL ➡️ Clean Text ➡️ Chunks ➡️ Vectors ➡️ Chat.

This allows users to point the app at their documentation and ask questions immediately.

🎁 Want the Full Source Code?

I cleaned up this entire logic (plus Multi-File PDF support, Mobile UI, and Streaming response) and packaged it into a production-ready Starter Kit called FastRAG.

It saves you the ~40 hours of setting up the boilerplate so you can focus on building your AI SaaS over the holiday break. 🎅

🏁 I'm running a "Holiday Build" race:

🥇 First 69 devs: Get 69% OFF (~$9). Code: FAST69

🥈 Everyone else: Get 40% OFF. Code: HOLIDAY40

Check out the Live Demo & Repo here: 👉 FastRAG

Happy coding and happy holidays! 🎄🚀

🛑 Stop Re-Writing RAG Pipelines: My Next.js + Pinecone Architecture ⚡️

Atul Tripathi — Fri, 21 Nov 2025 17:23:16 +0000

Let's be real for a second. ☕️

I have built the "Chat with your PDF" feature for clients about five times in the last two months.

The frontend? Fun. streaming UI, Tailwind, those fancy typing effects... I love it. 🎨
The backend? Absolute headache. 🤯

Every single time, I found myself staring at VS Code, copying and pasting the same boring boilerplate to handle:

❌ The PDF Loader: Chunking text without breaking sentences mid-thought.
❌ The Embeddings: Batching data to OpenAI so I don't hit rate limits.
❌ The Vector Store: Upserting into Pinecone/Supabase without errors.
❌ The Context Window: Calculating tokens so the AI doesn't crash.

After the 5th time, I realized I was wasting 40+ hours per project just setting up the "plumbing" before I could actually build the cool stuff.

So, I decided to fix it. Forever. 🛠️

🏗️ The Architecture

Here is the stack I finally settled on for a production-ready RAG (Retrieval Augmented Generation) app. It's typed, it scales, and it just works.

▲ Framework: Next.js (App Router)
🦜 Orchestration: LangChain.js
🌲 Vector DB: Pinecone (My go-to for speed)
⚡️ Database/Auth: Supabase
💅 Styling: Tailwind CSS

🧠 The Hard Part: Handling Vectors

The biggest pain point isn't the chat; it's the Ingestion Pipeline. You can't just dump a PDF into ChatGPT. You have to slice and dice it first. 🔪

Here is the logic I standardized:

User uploads file 📂
Server reads buffer 👓
RecursiveCharacterTextSplitter breaks it into chunks 🧩
Upsert to Pinecone with metadata 💾

It looks something like this (simplified for sanity):

import { OpenAIEmbeddings } from "langchain/embeddings/openai";
import { PineconeStore } from "langchain/vectorstores/pinecone";
import { RecursiveCharacterTextSplitter } from "langchain/text_splitter";

// 🪄 The logic that usually takes 3 hours to debug
export async function addDocumentsToVectorStore(text, fileId) {

  // 1. Split the text intelligently
  const splitter = new RecursiveCharacterTextSplitter({
    chunkSize: 1000,
    chunkOverlap: 200,
  });

  const docs = await splitter.createDocuments([text]);

  // 2. Embed and Upsert to the Cloud ☁️
  const embeddings = new OpenAIEmbeddings();

  await PineconeStore.fromDocuments(docs, embeddings, {
    pineconeIndex: index,
    namespace: `file_${fileId}`, // 🔒 Isolate vectors per file
  });

  console.log("Vectors Upserted. We are ready to chat. ⚡️");
}

📦 So, I Productized It.

I got tired of setting this up from scratch.

I took my personal repo, cleaned it up, added a polished UI, and turned it into a reusable Starter Kit.

What's inside the box?

✅ Pre-configured LangChain Setup

✅ Built-in Vector Ingestion (PDF/TXT/MD)

✅ Streaming Chat Components (ChatGPT style)

✅ Rate Limiting (Save your API credits!)

🧪 The Experiment (Smoke Test)

I'm running a little experiment this weekend.

Instead of charging the full launch price ($149), I set up a $9 Early Bird Deposit.

Why $9? It acts as a "Skin in the Game" filter. If the pain of building RAG pipelines isn't worth the price of a coffee ☕️, then I know this isn't worth building further.

But if you want to save 40 hours of work?

👉 Grab the RAG Starter Kit (Early Access)

(Even if you don't buy it, feel free to roast my landing page in the comments! I'm building this in public and feedback is gold. 🥇)

Happy Coding! 🚀

Yo!

Atul Tripathi — Mon, 03 Nov 2025 06:32:40 +0000

Data Structure and Algorithms roadmap (with resources)

Atul Tripathi — Mon, 03 Nov 2025 06:19:30 +0000

Atul Tripathi

Nov 1 '25

🚀 LeetCode in 2025

#leetcode #dsa #algorithms #coding

Comments

2 min read

🚀 LeetCode in 2025

Atul Tripathi — Sat, 01 Nov 2025 15:57:35 +0000

🚀 LeetCode in 2025

Big tech (Amazon, Google, Microsoft) still use LeetCode-style interview questions.
Over 3,000 problems make it overwhelming for beginners.
This guide aims to make your LeetCode journey simpler and less painful.

💼 Why LeetCode Matters

Not all companies ask DSA questions — startups often prefer project-based experience.
Big tech → LeetCode prep is essential.

💻 Programming Language Choice

Language doesn’t matter much — focus on problem-solving.
Python is beginner-friendly; stick to what you know (C++, Java, JS, Go, etc.).
Know the basics: variables, loops, conditions, arrays, functions, classes, I/O.
Learn built-in libraries to save time (Python: list, dict, set; Java: HashMap, etc.).

📚 Foundation Before Starting

Understand Big O, time complexity, and key data structures & algorithms.
Focus on arrays, strings, linked lists, stacks, queues, hash tables, trees.
Learn recursion — crucial for tree and graph problems.
Don’t chase perfection — just start learning.

🎓 Recommended Resources

Data Structures → WilliamFiset (YouTube)
Algorithms → Abdul Bari (YouTube)
DSA Patterns → AlgoMasterIO (YouTube)

🧩 Step-by-Step Learning

Learn one topic at a time (arrays → strings → linked lists → trees).
Solve 4–5 easy problems per topic to build confidence.

🧠 Smart Problem Selection

Start with easy problems, then move to medium.
Focus on Top 100 Liked & Top Interview 150 lists.
Aim for ~300 high-quality problems covering major patterns.
Use algomaster.io for a curated 300-problem roadmap.

🔍 Quality Over Quantity

Deeply understand each problem instead of rushing.
After solving, ask: Why does this work? What’s the key insight?

🧠 Pattern Recognition

Focus on problem patterns, not memorizing questions.
Learn key patterns (e.g., monotonic stack) and apply them across problems.
Find categorized pattern lists on algomaster.io.

✏️ Problem-Solving Strategy

Read carefully — note constraints & examples.
Start with brute force, then optimize.
Use prefix sums, hash maps, two-pointers, etc.
Always analyze time & space complexity.

⏱️ Practice Routine

Easy → 10–15 mins
Medium → 30 mins
Hard → 60 mins
Join LeetCode contests or simulate past ones to build speed.

🔁 Revision & Retention

Regularly revisit old problems to strengthen memory.
Rewrite solutions without looking.
Avoid memorizing — understand logic deeply.

💪 Progress Mindset

Improvement takes time and consistency.
Feeling stuck is normal — take breaks and return later.
Focus on learning, not perfection.

Build your own JARVIS 'Gem' 💎 on Gemini! 🤖 Prompt starter: ✨ "You are JARVIS, a witty & sophisticated AI 🧐. Your goal is to provide insightful solutions.💡" Customize it ⚙️ to create your perfect AI sidekick! #AI #Gemini #PromptEngineering

Atul Tripathi — Mon, 13 Oct 2025 03:56:13 +0000