WORKLINE AI：整合多個 AI 智能體的工程工作台

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

WORKLINE AI is an AI-powered engineering workbench that helps people move from an idea to a validated, documented, procurement-ready engineering project without manually coordinating every step.

Without WORKLINE AI

-> Manually searching research papers and datasheets

-> Comparing components across vendors

-> Repeatedly documenting requirements and architecture
-> Manually tracking dependencies and project progress

-> Repeating research whenever project context is lost

-> Switching between tools and coordinating agents manually
-> Rebuilding context across sessions or machines

With WORKLINE AI
-> AI-assisted research and knowledge extraction
-> Organized BOMs, component comparisons, and sourcing workflows
-> Structured, reusable engineering artifacts
-> Connected project data, tasks, and collaboration
-> .wl project intelligence helps the system understand where to start
-> A unified workbench with delegated, permission-controlled agents
-> Portable project data, backup, restoration, and synchronization

Who gets to be a little lazier?
-> Student engineers: Less time formatting reports and hunting for components; more time building prototypes.
-> Hardware engineers: Less repetitive documentation, component research, and cross-referencing.
-> Researchers: Less manual literature collection and synthesis.
-> Startup founders and small teams: Less project coordination and administrative overhead.
-> Engineering leads: Less chasing updates, reviewing disconnected artifacts, and tracking dependencies.

WORKLINE does not eliminate engineering responsibility. It reduces the coordination tax around engineering: repetitive research, documentation, context reconstruction, and tool switching.

The goal is simple: automate the busywork so engineers can spend more time thinking, designing, testing, and building.

Demo

Code

📌 Executive Summary

Modern hardware engineering is fragmented across disparate silos: schematic capture, distributor catalog queries, thermal/electrical physics simulations, microcontroller firmware authoring, and procurement logistics.

Workline solves this through a unified Agent Control Fabric powering 27 specialized Google ADK domain agents across 90+ capabilities, paired with a Local Engineering Intelligence Runtime:

1. Requirements Decomposition: Translates plain text into structured mechanical, electrical, thermal, and regulatory constraints.
2. Local-First .wl Project Model…

How I Built It

-> Local vector database
Qdrant
Stores vector embeddings for semantic project retrieval.
-> Local retrieval

Moss
Supports local semantic/hybrid retrieval and project-context discovery. -> Persistent project database

SurrealDB
Stores structured engineering data, relationships, and project knowledge.
-> Agent development
Google ADK
Supports building and orchestrating AI agents.
-> Agent interoperability
A2A + Bindu
Supports communication and interoperability between agents and external systems.
-> Model inference
NVIDIA MODEL Library
Provides configurable model inference and reasoning.
-> Agent execution
Local agents + optional Agent37
Runs local tasks or delegates browser-based, research, and isolated execution tasks to a sandbox.
_-> Governance

ArmorIQ
_Supports scoped delegation, authorization, and auditable agent actions.

Why Does Open Innovation Matter?

Open innovation made WORKLINE AI modular, extensible, and less dependent on a single AI provider. Instead of building everything around one closed API, I designed it to combine open frameworks, local infrastructure, and configurable model providers.
-> Local-first engineering: Qdrant, Moss, and SurrealDB support local retrieval and project data management, reducing dependence on external services for core project context.
-> Interoperable agents: Google ADK, A2A, and MCP provide building blocks for coordinating agents and connecting tools instead of locking every workflow into one proprietary ecosystem.
-> Flexible execution: Local agents can handle suitable tasks, while Agent37 can provide isolated execution for browser research and other delegated work.
-> Governance and accountability: ArmorIQ can help enforce delegation boundaries and trace agent actions across workflows.
-> Provider flexibility: The architecture can accommodate different model providers without redesigning the entire engineering workbench.

A closed API could enable powerful reasoning, but it wouldn't automatically provide this level of control over local execution, retrieval, agent interoperability, and deployment.

The key advantage of open innovation is architectural freedom: WORKLINE can assemble, replace, and evolve individual components without rebuilding the entire platform.

My Agent Session

[ agent_session 1096 ]

The Lazy Payoff

WORKLINE is designed to reduce repetitive engineering work, but we would need real task timings and usage data to quantify the benefit reliably.

   Engineering task       Manual workflow    Potential with WORKLINE
-> Collecting and organizing   2–4 hours            30–90 minutes
   research
-> Preparing a preliminary     1–3 hours            20–60 minutes
   BOM
-> Creating initial project    2–4 hours            30–90 minutes
   documentation
-> Reconstructing project     30–60 minutes       5–15 minutes
   context
-> Coordinating tasks and     1–2 hours/week      20–60 minutes/week
   engineering artifacts

Prize Categories

-> Render
-> Gemma

Team Submissions: devgenius (Team Lead and developer)