Open Source & Engineering

Frameworks and tools I build

From ML frameworks to agent orchestration engines and reference architectures — built to test ideas in practice, solve recurring engineering challenges, and accelerate production delivery.

These are my own projects, built and maintained under UNICOLAB. Where a project builds on third-party work, the card says so.

Projects: 15 · Public source: 12

Agent Frameworks

04

  • No screenshot yet
    ActiveAgent frameworkPythonMIT

    Agentomatic

    Drop agents, not code. A zero-code multi-agent API framework: it discovers agents from a folder and exposes them as REST endpoints, with auto-routing, streaming, the A2A protocol, tracing and time-travel debugging.

    PythonFastAPIA2AStreaming+1
    Install
    pip install agentomatic
  • GoLangGraph — screenshot
    ActiveAgent frameworkGoMIT

    GoLangGraph

    Go framework for building AI agent workflows using graph-based execution at native Go performance. Define nodes, edges, and conditional routing to orchestrate complex LLM pipelines.

    A Go implementation of the LangGraph framework

    GoLangGraphGraph DAGLLM
    Install
    go get github.com/UnicoLab/GoLangGraph
  • GoLangGraph Studio — screenshot
    ActiveAgent frameworkReact

    GoLangGraph Studio

    React-based debugging IDE for testing and visualizing GoLangGraph agent flows. Step-through execution, breakpoints, and real-time state inspection.

    Adapted from LangGraph Studio

    ReactTypeScriptWebSocketGo
  • No screenshot yet
    BetaAgent frameworkGo

    AgentOS

    Enterprise-grade agent operating system with modular domain-specific packs, RBAC governance, and on-premise deployment architecture.

    GoPythonKubernetesRBAC

ML Frameworks

05

  • KerasFactory — screenshot
    ActiveML frameworkPythonMIT

    KerasFactory

    Reusable Keras layers and model templates for tabular data processing and feature engineering. Rapid model prototyping with proven production architectures.

    Built on Keras

    PythonTensorFlowKeras
    Install
    pip install kerasfactory
  • MLPotion — screenshot
    ActiveML frameworkPythonMIT

    MLPotion

    Building blocks for customized ML pipelines — atoms, steps, and components for rapid experimentation. Compose complex workflows from simple, tested primitives.

    PythonScikit-learnPipelineTesting
    Install
    pip install mlpotion
  • FlowyML — screenshot
    ActiveML frameworkPythonApache-2.0

    FlowyML

    Comprehensive ML pipeline framework combining development simplicity with enterprise MLOps power. Used by 30+ teams for production model delivery.

    PythonMLOpsPipelineCI/CD
    Install
    pip install flowyml
  • No screenshot yet
    BetaML frameworkPythonMIT

    GraphFlow

    Graph-based data processing framework with automatic context management: define processing steps as a graph and let the framework carry context between them.

    PythonGraphsData processing

Tooling

03

  • No screenshot yet
    ActiveToolingPythonApache-2.0

    FlowyML Notebook

    Interactive ML notebook environment integrating KDP, KerasFactory & MLPotion for rapid prototyping. The complete ML ecosystem in one tool.

    PythonJupyterKDPKerasFactory
    Install
    pip install flowyml-notebook
  • No screenshot yet
    ActiveToolingPythonMIT

    HolySheet

    A Python-first report compiler that generates self-contained, interactive HTML dashboards powered by React — without the end user installing Node.js, npm or any frontend tooling.

    PythonReactTypeScriptDashboards
    Install
    pip install holysheet
  • No screenshot yet
    BetaToolingGoMIT

    smlcode

    A small but capable coding harness optimised for small language models (SLMs), written in Go.

    GoSLMsCLI
    Install
    brew install --formula https://raw.githubusercontent.com/UnicoLab/smlcode/main/Formula/slmcode.rb

Reference Platforms

The source for these platforms is not public. They are shown as reference architectures.

03

  • No screenshot yet
    ActiveReference platformGo

    AIFlow Architecture

    End-to-end AI workflow and agent orchestration platform with flow debugging and state inspection — engineered with Go, React, and Kubernetes.

    GoReactTypeScriptKubernetes+1
    Source not public
  • Watchtower Hub — screenshot
    ActiveReference platformGo

    Watchtower Hub

    Centralized licensing, entitlements, and runtime observability system featuring multi-tenant isolation, Redis-backed seat leasing, and runtime validation.

    GoReactRedisHMAC/JWKS+1
    Source not public
  • Galaxy Architecture Hub — screenshot
    ActiveReference platformGo

    Galaxy Architecture Hub

    Centralized engineering repository and architecture hub for structuring, evaluating, and sharing reusable AI/ML system blueprints across teams.

    GoReactTypeScriptPostgreSQL+1
    Source not public