Project lab

Public projects behind the AI news and agent workflow stack.

This index maps my GitHub repositories and production systems into a readable portfolio: what each project does, why it exists, and how it connects to mlllm.io.

Projects are the credibility layer of mlllm.io. They show that the news desk and the writing are backed by real software work: publication pipelines, agent planning dashboards, MCP servers, local QGIS workflows, video-editor copilot experiments, Telegram automation, and model-deployment practice.

The list is based on the public GitHub profile plus the production TG-NEWS system that powers the site. The profile README repository itself is identity metadata, so it is not listed as a separate project.

Core systems

Production and agent-planning work.

Production

TG-NEWS

The upstream AI news pipeline behind mlllm.io and the Telegram channel. It collects public AI signals, maps them into story records, prepares short and expanded publication payloads, and feeds the static site archive.

AI newsTelegramSite pipeline

Public OSS

Task-Plan v2 Dashboard

A dashboard and documentation package for structured multi-agent task plans: implementation gates, review state, execution visibility, examples, prompts, and dashboard-oriented planning artifacts.

JavaScriptPlanning systems

AgentsDashboardTask plans

Public OSS

Mouse Trail Masking Reveal

A two-skill Agent Skills workflow for validating image-layer alignment and building a canvas-based soft cursor-trail reveal hero with tuning controls and browser smoke checks.

JavaScriptCanvas UX

Agent SkillsCanvasReveal

Experimental

Auto-iFlow

An autonomous multi-agent coding framework based on the now unsupported iFlow line. It is useful as an experimental reference for planning, building, validating, and visualizing coding work with a Kanban-style workflow.

TypeScriptMulti-agent coding

Auto-iFlowKanbanAgents

Experimental

AFOC

Auto Flow Open Code: an Auto-iFlow experimental repository focused on OpenCode CLI integration. It sits between the older Auto-iFlow workflow and newer agent execution surfaces.

TypeScriptOpenCode integration

OpenCodeAuto-iFlowCLI

Editor and GIS tools

Agent workflows inside real applications.

Public OSS

OpenCut OpenCode Copilot

A publication-ready OpenCut fork that adds an OpenCode-style AI control layer for video editing: session APIs, model lists, health checks, chat messages, action planning, and an editor copilot surface.

TypeScriptVideo editor

OpenCutAI copilotSemantic video

Public OSS

MCP-QGIS

An MCP/QGIS bridge for LLM-driven geospatial workflows. It exposes QGIS-related operations through a local MCP service and is focused on practical GIS and cadastral scenarios.

PythonQGIS bridge

MCPQGISGeospatial

Prototype

QGIS-OC-CHAT

A companion QGIS plugin and OpenCode sidecar prototype for chat-driven GIS workflows. It explores human-in-the-loop confirmation, RAG context, and safer local tool execution.

PythonQGIS plugin

PluginSidecarHITL

MCP integrations

Reusable agent interfaces for external services.

Public OSS

45telega

A production-oriented Telegram MCP server with broad Telegram method coverage, deployment paths, and Claude/agent integration use cases. It is the Telegram automation side of the portfolio.

PythonTelegram MCP

TelegramMCPAutomation

Public OSS

mlllm-threads-mcp

A TypeScript MCP server for Meta Threads with post creation, media-oriented workflows, analytics, conversation management, and social publishing support for LLM-generated content.

TypeScriptThreads MCP

ThreadsSocialPublishing

Public OSS

Perplexity Pro MCP

A high-performance MCP server for Perplexity API workflows, including search, multi-search, deep research, fact checking, code search, academic search, and recent-search patterns.

MCPResearch search

PerplexityResearchFact check

Fork

threads-mcp

A fork of a Threads MCP server pattern. It is useful as a reference point for OAuth, media uploads, analytics, and enterprise-style Threads API integration.

ShellFork

ThreadsOAuthReference

Model and deployment experiments

Forks used for learning deployment and AI media infrastructure.

Fork

DeepSeek R1 Distill Qwen 32B deployment

An Inferless/vLLM deployment template for DeepSeek-R1-Distill-Qwen-32B. It documents runtime configuration, model initialization, inference, and deployment testing.

vLLMInferless

DeepSeekQwenDeployment

Fork

Qwen3 14B deployment

A model deployment template for Qwen3-14B with vLLM-style serving and Inferless configuration. It is part of the practical model operations layer behind AI workflow experimentation.

vLLMModel serving

Qwen3LLMServing

Fork

Stable Video Diffusion deployment

A deployment template for image-to-video generation with Stable Video Diffusion and Diffusers. It supports the media-generation side of the broader AI tooling portfolio.

DiffusersVideo generation

SVDVideoInferless

Fork

Wan2.1

A fork of the Wan2.1 open video generation model repository. It is not presented as original model research; it is a working reference for video generation experiments and model operations.

PythonVideo model

WanVideoFork

Template

Inferless onboarding

A compact Inferless onboarding template that shows how to structure `app.py`, request/response schemas, and deployment-ready model code for GitHub-based deployment.

InferlessTemplate

DeploymentTemplateModel ops

Source

Repository facts are drawn from the public GitHub profile and repository README files. The profile README repository is treated as identity metadata, not as a software project.

Open GitHub profile Author profile