SShortSingh.
Back to feed

Developer builds multi-agent AI fitness planner in TypeScript using HazelJS framework

0
·1 views

A developer has detailed how to build an intelligent fitness workout planner using HazelJS, a TypeScript framework designed for constructing multi-agent AI systems. The solution uses five specialized agents — covering intake, exercise search, workout planning, progress tracking, and coaching — each handling a distinct part of the fitness planning process. Unlike traditional fitness apps that rely on static templates, the agentic approach allows for personalized, adaptive plans based on user goals, fitness level, available equipment, and injury history. HazelJS features such as the @Agent, @Tool, and @Delegate decorators enable agents to collaborate and delegate tasks without tightly coupling their logic. The modular design means individual agents can be improved independently, making the system easier to maintain and scale over time.

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)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

How Vision Language Models Taught AI to See and Understand the World

Vision Language Models (VLMs) combine image understanding and natural language processing into unified AI systems capable of describing, reasoning about, and answering questions on visual content. Early approaches paired CNN encoders with RNN decoders to generate basic image captions, but these systems lacked true scene comprehension. OpenAI's CLIP marked a turning point by aligning images and text in a shared embedding space, enabling zero-shot visual recognition without explicit task training. Models such as Flamingo, BLIP-2, and LLaVA extended this by generating free-form conversational responses about images. Today's frontier systems — including GPT-4V, Gemini, and Claude — are built from the ground up as multimodal architectures, treating text, images, video, and audio as native inputs rather than bolted-on additions.

0
ProgrammingDEV Community ·

Deterministic guardrails can stop AI agents from making dangerous mistakes

AI agents equipped with real-world tools like package managers, wallets, and email accounts can autonomously perform harmful actions such as installing malware-laced packages, executing prompt injections, or sending payments to sanctioned addresses. Using a second AI model to review outputs is unreliable because it adds latency and can itself be manipulated by the same injection attacks it is meant to catch. A more effective approach uses deterministic, rule-based checks that perform a single factual lookup with no model inference, returning a consistent verdict in milliseconds. A set of free APIs has been developed to handle common risk categories including package verification, content scanning, code analysis, and payment screening, each returning a simple allow, review, or block verdict. These guards can also be integrated directly into MCP-compatible AI coding tools, making them a low-friction pre-step before any consequential agent action.

0
ProgrammingDEV Community ·

How One Developer Turned LinkedIn Frustration Into a Meme-Making Journey

A developer frustrated with LinkedIn's highlight-reel culture decided to inject humor into the platform by posting tech-related memes. Early attempts flopped, drawing almost no engagement, but participation in Dev.to's weekly Meme Monday thread provided encouragement when a post received a reaction from the community's founder. Undeterred by low like counts on a JavaScript LinkedIn group, the developer kept experimenting and refining their meme-making skills. Persistence eventually paid off when an April Fool's Day meme crossed the 100-likes milestone for the first time. The journey reflects how developer humor and online community engagement can turn professional frustration into a creative outlet.

0
ProgrammingDEV Community ·

How Vision Transformers Replaced CNNs by Treating Images as Patch Sequences

Vision Transformers (ViT) adapt the Transformer architecture — originally designed for text — to process images by dividing them into fixed-size patches treated as sequential tokens. This approach was made possible after Transformers revolutionized NLP following Google's 2017 'Attention Is All You Need' paper, which introduced self-attention mechanisms. For decades, Convolutional Neural Networks (CNNs) dominated computer vision by using learnable filters to detect local patterns hierarchically, but they struggled with long-range dependencies and rigid geometric processing. ViT addresses these limitations by allowing every image patch to attend to every other patch, enabling richer global context understanding. A standard ViT splits a 224x224 image into 196 non-overlapping 16x16 patches, each flattened and projected into a token vector before being fed into the Transformer.