2026 AI toolkit

Here's what I'm actually using right now.

Inference

• Ollama — runs models, downloads models, serves models
• llama.cpp — the engine under Ollama

Models

• huihui_ai Qwen models — abliterated, general purpose
• Qwen2-VL-7B-Instruct-abliterated — vision tasks
• nomic-embed-text — embeddings for RAG

Databases

• pgvector — vector search in Postgres
• Apache AGE — graph database in Postgres
• Dgraph — standalone graph database

Image and audio

• stable-diffusion-webui — SDXL image generation
• ComfyUI — for agentic image pipelines (API only)
• NeuTTS Air — text-to-speech

Agent infrastructure

• MCP TypeScript SDK
• MCP Go SDK
• Temporal — agent run loops and workflows
• Docker — isolated agent runners

Coding

• Cursor — AI coding assistant
• Eleventy — static site generator

Backend / infra

• PocketBase — backend in a binary

Notes and knowledge

• Joplin — notes, with my own MCP server

Browser automation

• Puppeteer — headless Chrome

On the radar

• Unsloth — quantized LoRAs
• SGLang — inference engine
• LibreChat — multi-provider chat interface