Projects
The open-source tools I build and teach with: AI Agents, Local Models, and AI Workflows. Most of them started as a livestream. Clone the code on GitHub or try the live versions.
AI Agents
Orchestrators, desktop agents, and the tooling I use to watch and extend them.
Codex Astra/Luna Orchestrator
Installable Codex profiles that use GPT-6 Astra or GPT-6.1 Sol as the orchestrator, GPT-6 Luna subagents to explore, build, research, and test, and an independent reviewer.
Coworker
A local-first desktop app for AI coworkers on macOS, Windows, and Linux, with no subscription. Use local models through Ollama or LM Studio or bring your own keys, and get skills, memory, a scheduler, approvals, and Telegram or Discord.
Agent Monitor
A local dashboard for Codex, Claude Code, Pi, and Hermes. Explore agent trees, read traces, and track token usage and estimated costs. Traces stay on your machine.
Skillsbento
Free agent skills packaged as plugins you can install in Claude Code and Codex, from stock analysis and sales or social media dashboards to product launch docs and livestream-to-shorts.
GAIA Personal Setup
An installer for running several private Hermes Agent assistants side by side on your Mac or Windows PC. Each one gets its own Docker container, data folder, API key, and Telegram or Discord bot.
Hermes Profiles
Installable Hermes Agent profiles for a legal assistant, real-estate agent, home manager, and content creator. Each one ships its own SOUL, config, skills, and cron templates.
Cursor Go SDK
An unofficial Go SDK for the Cursor Cloud Agents API: typed REST methods, SSE run streaming, artifacts, and a higher-level Agent/Run interface inspired by the TypeScript SDK.
Local Models
Run, evaluate, and work with open models on your own hardware.
Local Evals
A local-first app for evaluating document-to-JSON, text-to-JSON, and tool-call outputs with local models, any OpenAI-compatible API, or OpenRouter. Inputs, outputs, scores, and logs sit side by side, and there is a CLI too.
AIBackends (Python)
A Python library for running AI tasks and workflows locally with llama.cpp and Transformers: extraction, classification, embeddings, PII redaction, moderation, and prompt routing.
Open Model Server
A self-hosted server for local GGUF and ONNX models. One Go binary provides the HTTP API, a web dashboard, a SQLite model registry, and model lifecycle management.
JSONL Viewer
A desktop app for viewing and inspecting .jsonl / .ndjson files, the usual format for fine-tuning datasets. It also opens agent traces from Codex, Claude Code, Pi, and Hermes.
Markdown Editor
A lightweight Markdown editor for Windows, Mac, and Linux with live preview, Mermaid diagrams, and AI editing that runs on local models through Ollama or LM Studio.
AI Workflows
APIs and pipelines that wire AI into apps and business systems.
AI Backends API server
An API server for adding AI to your apps, run locally or self-hosted. Ready-made endpoints for text, vision, and data tasks, plus tool-using agents, on Ollama, LM Studio, or llama.cpp, or on OpenRouter, OpenAI, Anthropic, and Google.
CamelFlow
A viewer for Apache Camel YAML routes. It shows your routes and how they connect as an easy-to-read diagram and refreshes every time you save.
Learn with code
Notebooks and examples I share with students and viewers.
Notebooks
Local model fine-tuning and inference notebooks you can run in Google Colab, including QLoRA fine-tunes of Gemma 4 E2B with Unsloth, GLiNER2.5 and LFM2.5 examples, and AIBackends walkthroughs.
Learn Go
The Go code I wrote while learning Go, now with Generative AI examples such as translating text with Groq's API.
Star counts from GitHub as of Oct 6, 2026.See all repos on github.com/donvito