DNA Platform - Bioinformatic Workflow Engine

A low-code/no-code platform for computational biology workflows using .NET agents framework with drag-and-drop workflow designer.

🎯 Overview

DNA Platform enables bioinformaticians to:

• Build complex workflows DSL using a visual WYSIWYG tooling drag-and-drop reusable commponet
• Centralized polyglot runtime one monolithic platform, automated dependency resolver
• Compose reusable nodes plug and play, validate scheme before running the workflow
• annotate any library use plugin MCP servers or run locally via predefined serializied objects [smart_commponets]
• Interoperate with Python, R, Node.js, JVM or .net core all through polyglot platform ready to scale

Built on: .NET 8.0+ and Node LTS | uses Mise MCP poyglot resolver for workflow or platform extnesions tools. Philosophy: Polyglot backend :dotNET core with CSsnake runtime, sveltekit for Frontend. middlewareAPI proxy with seperation of concerns and DRY princples. Author workflow with AG-UI Protocol integrate A2A,ACP, FFI, IPC, and FaaS for durpable cloud-native scalability. Use CLI with python, bash or PowerShell_Core for Cmdlet,Structured .NET objects and extendability e.g Rapid tool development and Automation.

✨ Key Features

Visual Workflow Canvas: Build and test directed acyclic graph (DAG) pipelines via an intuitive drag-and-drop designer. Polyglot Execution with CSnakes: Execute compiled C# .NET algorithms and Python bioinformatics packages (BioPython, NumPy) side-by-side with zero-overhead data transfer. Human-in-the-Loop (HITL) Governance: AI agents stream reasoning traces and pause automatically before sensitive operations (e.g., sequence mutations, system execution) to request operator sign-off.

🔧 Agent-Driven Workflow Engine

:

• Topological execution for correct dependency ordering
• Parallel node execution for performance
• Smart retry policies with exponential backoff
• Full async/await support for non-blocking operations
• use agents as MCP service use or build agents alongside with tools and skills_set .
• recompose any workflow preset as nested node use any prebuilt workflow
• CLI moudles Audit CLI tools and extensions (WIP)
• Human-in-the-Loop AI agents stream reasoning traces and pause automatically before sensitive operations.
• RLHF & MARL model native paradigm with lightning and meta MARL (WIP)
• Statemachine workflow a tab for seperate complex pipelines

🎨 Extensible Node Types

• strongly typed Input/Output autoboxed if needed
• Agent nodes with AI reasoning capabilities
• Processing skills for deterministic FILE system operations e.g save or load or marshling data or pip objects.
• Conditional branching for complex logic
• Loop constructs for iteration
• Subworkflow references for composition
• CLI node for exposing stdio machansim
• CRON JOBS for automated JOBS

🤖 AI Agent Framework

• Provider-agnostic IAgent — Microsoft Agent Framework (hosted),A2A with AG-UI Protocol
• Local LLM inference support (onnxtuntime) (ollama wip)
• Tool calling for local skill invocation
• Reasoning traces for interpretability (ACP protocol stud IO observability layer)
• Mutlomodal support mulitple data formats

🧩 Extensible MCP & Skill System

• Plugin architecture for customizing complex bioinformatic pipelines locally
• Built-in skills: Basic Bioinformatics tool sample, Data transform, Validation, Merge results
• Dynamic skill discovery and registration (test)
• Modular Modular BCL&FCL DI pattren scaffold prewritten library into Boilerplate Code (stubs) and S you can also use CLI's future for automated insalls and annotation
• Input/output schema params testing and validation
• MCP and tools tests on the fly
• Marketplace and ecosystem local or remote MCP filter agents, mcp tools ,moudles, extensions etc..

🌐 Polyglot Runtime

Polyglot Execution with CSnakes: Execute compiled C# .NET algorithms and Python bioinformatics packages (BioPython, NumPy) side-by-side with zero-overhead data transfer.

• .NET polyglot kernel for Jupyter notebook connected to the platform
• Python interop (with csnake runtime)
• Node.js bridge for JavaScript tooling e.g threejs and D3 for interactive widgets or alternative scripting
• mise-based environment — mise.toml pins the toolchain and test for workflow needed env in the ⚙️ Configure view.
• pack drylab experiments make an abstract interactive webapps sort of like hugging face's playground locally for now maybe -----(WIP:Distroless docker image) Dual Interfaces: Work seamlessly in your browser via the SvelteKit web studio or in the terminal using the native Terminal.Gui console dashboard.

💻 DevUI - Web-Based Designer

• Drag-and-drop workflow builder WYSIWYG for composing nodes and reusable commponet on a canvas
• Workflow persistence via Prisma/SQLite (SvelteKit server routes)
• Backend execution most endpoints proxied to the .NET orchestrator
• REST API CRUD for workflow management, agents etc..
• Stateless by deafult Native implementation of the AG-UI Protocol (docs.ag-ui.com).
• Real-time execution monitoring for stdio/sse transport, tool calling etc..
• Customizble

🚀 Quick Start (DevUI Web App)

Prerequisites

• .NET 8.0+ SDK installed (net8.0)
• Node.js 20+ installed
• Python 3.10+ (optional, for CSnakes BioPython/NumPy interop; pure C# fallback included)
• mise (optional, recommended for toolchain pinning)

The 3 Core Scripts (Cross-Platform)

1. One-Command Setup & Bootstrap

Probes your toolchain, configures .env, builds the C# Terminal.Gui CLI, and syncs the Prisma SQLite database:

npm run bootstrap
# or: node scripts/1-bootstrap.mjs

2. Run the System (DeNovoAutomata)

Boots instantly in under 1 second without reinstalling. Starts the Web Studio, ASP.NET Core API, and AG-UI server with an interactive terminal controller:

npm start
# or: node scripts/3-denovo-automata.mjs
# or on Windows: run-all.bat

This starts:

• Web Studio (DevUI): http://localhost:3000 (or http://localhost:5173)
• Backend API: http://localhost:5254 (Swagger: /swagger)
• AG-UI Protocol Server: http://localhost:5000 (/api/agui)

3. Test & Clean

Runs the automated 10-point agentic E2E verification suite. Use test:clean to clean temporary files and caches before Git push:

# Run test suite
npm test
# or: node scripts/2-test-dev.mjs

# Run test suite and clean workspace artifacts
npm run test:clean
# or: node scripts/2-test-dev.mjs --clean

4. Interactive Terminal.Gui Console TUI

Launch the native windowed C# terminal interface:

npm run tui
# or: node scripts/dna.js tui

Optional: Pin Toolchain with mise

mise install        # installs the dotnet 8 + node lts versions mise.toml pins
mise tasks ls       # confirms build, build:web, dev were picked up
mise run build      # dotnet build DNAPlatform.sln

---

📚 Documentation

• System Architecture (docs/architecture.md) — Low-code/no-code bioinformatic platform architecture, folder structure, agentic development guide, troubleshooting, and access points.
• Development Guide (docs/development.md) — Merged AGENTS guide, command line usage, completed milestones, and roadmap.

---

📄 License

MIT License - Open source for research & development

---

Status: 🚀 Alpha | Version: 0.3.0-alpha | Updated: 2026-10-17