AttentionSeekers

Attention Seekers: Agentic Voice-to-CAD System
Project Description for Hackathon Submission
Core Functionality

Attention Seekers is an advanced agentic speech-to-CAD automation system that converts natural language design requests into manufacturable 3D models. The system uses a coordinated team of specialized AI agents, each responsible for a different phase of engineering design. A Requirements Engineer converts voice or text into precise specifications with automatic clarification. A physicist agent — powered through an external N8N microservice that performs real-time web-assisted analysis — validates physical feasibility and manufacturability. A CAD Coder generates clean FreeCAD Python scripts, a Reviewer checks for dimensional and functional correctness, and a Packager outputs ready-to-print STL files, specification sheets, and concise voice summaries.

The pipeline supports iterative refinement, looping back when physics blockers or reviewer issues arise, enabling high-fidelity designs even from ambiguous initial inputs. The architecture is fully modular, allowing agents to run independently or in parallel, depending on computational and workflow requirements.

Working Prototype Stability

The system demonstrates robust production-grade behavior, with strict validation at every stage.
Key stability characteristics include:

Spec completeness checks with safe fallback paths

N8N-based physics microservice with validation, error handling, and structured JSON guarantees

Automatic failure-safe loops when physics blockers, malformed code, or invalid geometries appear

CAD script syntax checking and semantic validation

Timeout and exception management across agent calls

Deterministic revision cycles that converge to manufacturable designs

Attention Seekers has been tested on a wide range of use cases — basic primitives, multi-feature parts, phone stands, brackets, wearables, and functional assemblies — demonstrating reliability across diverse geometries and materials.

Technical Complexity: Agent Orchestration + N8N Integration

The system exhibits significant technical complexity through multi-modal, multi-agent orchestration across cloud AI services, local tools, and an external N8N workflow.

Highlights:

Requirements → Physics → CAD → Reviewer → Packager orchestrated through LangGraph

An N8N Physics Feasibility Service acting as an autonomous micro-agent

Webhook ingest → query builder → web search → LLM physics evaluation → JSON output

Parallelizable branches (e.g., reviewer and packager after approval)

Tool invocation for FreeCAD, STL generation, and output bundling

Cross-agent memory, state tracking, and controlled revision loops

Seamless integration between cloud LLM reasoning and local CAD tool execution

This architecture demonstrates true agentic behavior, not simple prompt chaining.

Innovation & Creativity

Attention Seekers introduces a novel agentic approach to engineering design — turning natural conversation into manufacturable CAD output. Key innovations include:

A speech-driven design loop, enabling hands-free engineering workflows

Automatic translation from ambiguous human descriptions to precise 3D geometry

Real-time physics analysis integrated directly into the design pipeline

A structured, multi-agent architecture with well-defined roles and failure handling

Hybrid cloud + local engineering tool orchestration

Dynamic revision cycles where agents negotiate constraints until the design stabilizes

The result is a fluid design process where users can talk through ideas, receive engineering feedback instantly, and iterate without touching CAD software.

Real-World Impact

Attention Seekers lowers the barrier to 3D modeling and rapid prototyping by enabling everyday users — including hobbyists, students, makers, and entrepreneurs — to produce manufacturable CAD models using conversation alone.

Potential impact areas:

Rapid prototyping: cut iteration cycles from hours to minutes

Small manufacturers: generate custom parts without expensive engineering talent

Accessibility: empower users with mobility or visual limitations to design hands-free

Education: intuitive entry point for learning CAD and engineering concepts

Makers & hackers: instantly convert ideas into printable parts

Error prevention: integrated physics analysis avoids expensive print failures

By merging engineering expertise with conversational AI, the system makes CAD truly approachable.

Theme Alignment: Browsers, Voices, Clouds, and Tools as Agents

Attention Seekers embodies all four hackathon themes as active agentic components:

🌐 Browsers as Agents

A lightweight web or app interface becomes an interactive agent that:

Captures voice or text

Displays status of each agent

Presents 3D previews

Provides iterative feedback and clarification

🎤 Voices as Agents

Voice input becomes a design agent:

Accepting natural language descriptions

Delivering output summaries

Integrating with TTS for accessibility

☁️ Clouds as Agents

Cloud models (OpenAI, etc.) handle:

Requirements extraction

CAD code generation

Physics reasoning (with N8N orchestration)

Review and packaging logic

🛠️ Tools as Agents

Engineering tools operate autonomously:

FreeCAD for script execution

STL exporters

Physics calculators

N8N microservices as external tool agents

All four agent types collaborate in a unified, purpose-driven ecosystem designed to turn imagination into engineered reality.

Summary

Attention Seeker transforms engineering from a technical skill into a natural conversation. By integrating speech interfaces, multi-agent intelligence, N8N automation, cloud reasoning, and CAD tooling, it provides a complete end-to-end pipeline from voice prompt to manufacturable 3D model. The system demonstrates how cohesive, specialized agents can collaborate to solve traditionally complex tasks with simplicity, speed, and accessibility.

AI Engineers in startups