Tutorial: Build a production-ready AI agent using LangChain.js and NestJS

·2 views

A developer from SOM-OS has published a hands-on tutorial detailing how to integrate an AI agent into a NestJS application using LangChain.js. The architecture routes user requests through a NestJS controller into a BullMQ queue, where a worker processes each job asynchronously. A LangChain agent powered by GPT-4o handles the requests using Zod-typed tools, while conversation history is persisted in PostgreSQL and memory is cached via Redis. The guide includes full code snippets covering module setup, queue configuration, and agent execution. According to the author, this is the same architecture used in their own production systems rather than a purely theoretical exercise.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Cloudflare's Flexible SSL leaves your origin traffic unencrypted despite padlock

Cloudflare's Flexible SSL mode encrypts traffic only between visitors and Cloudflare, leaving the connection between Cloudflare and the origin server as plain HTTP. This means sensitive data such as session cookies and login credentials can be exposed in transit, even though browsers display the standard security padlock. The issue is widespread because the padlock gives users, developers, and security scanners a false sense of complete encryption. The recommended fix is to switch to Full (Strict) mode, which requires installing a valid certificate on the origin server — either via Let's Encrypt or Cloudflare's own Origin CA. Enabling the 'Always Use HTTPS' toggle in the Cloudflare dashboard is also advised to prevent downgrade attacks on the initial request.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why React and Next.js are often the wrong choice for marketing websites

Developers frequently use heavyweight JavaScript frameworks like React, Next.js, and Angular to build simple marketing websites, despite these tools being designed for complex, interactive applications. This mismatch results in large JS bundles, slower load times, and poor Lighthouse scores — directly hurting SEO and user retention. Google's research links load times beyond three seconds to sharply higher bounce rates, making performance a business-critical concern for lead-generation pages. Simpler alternatives such as plain HTML/CSS, WordPress, Webflow, or the static-first framework Astro typically deliver faster load times, better Core Web Vitals, and easier content editing for non-technical users. The core advice is to match the tool to the actual requirements of the project before choosing a technology stack.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

12 Beginner-Friendly Python AI Projects to Build Your ML Portfolio

A curated list of 12 Python-based AI starter projects has been outlined for beginners looking to enter machine learning without tackling deep learning immediately. The projects range from classic exercises like Iris flower classification and spam email detection to more applied tasks such as house price prediction, sentiment analysis, and a basic chatbot. Each project uses accessible tools including scikit-learn, pandas, numpy, and optionally Streamlit or LLM APIs from OpenAI and Hugging Face. Free datasets from Kaggle and scikit-learn's built-in collections are recommended as starting points, with Google Colab or Jupyter Notebook suggested as development environments. The guide advises beginners to complete one to three classic ML projects first before progressing to user interfaces and generative AI features.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Password Spray Attacks Bypass Lockouts and Compromise Enterprise Accounts

Password spraying is a cyberattack technique where criminals test a small set of commonly used passwords across a large number of user accounts, deliberately avoiding account lockout triggers. Unlike brute-force attacks that target a single account with many password attempts, spraying spreads low-frequency attempts across many accounts, making it far harder to detect. The method is particularly effective against cloud services, VPNs, Microsoft 365, and Single Sign-On platforms, which offer centralized access to multiple systems. Attackers exploit predictable human password habits — such as using seasons, company names, or keyboard patterns — and rely on publicly available lists of the most common passwords. Even a small success rate can be significant: in a 100,000-account organization, just 0.1% of weak passwords could hand attackers roughly 100 entry points into the network.

0 comments Read more at DEV Community