vLLM
High-throughput LLM serving with PagedAttention and an OpenAI-compatible API. The go-to for serving open models on a GPU server.
Links & repos
A curated, hand-picked list of links and GitHub repositories I find genuinely useful: AI agents, LLMs, self-hosting and my own projects. Stars and languages are refreshed at build time. Also available as JSON.
Find me online
30 items
Victor Grösser on Medium
My Medium profile, with cross-posts of articles from this blog on AI, agents and legal AI.
Uptime Kuma
Simple self-hosted uptime monitoring with notifications.
Unsloth
Fast, memory-efficient LoRA/QLoRA fine-tuning on a single GPU.
TRL
Hugging Face's library for SFT, DPO, GRPO and other post-training methods.
smolagents
Hugging Face's small, readable agent library with code-writing agents. Great for learning how agent loops actually work.
selfh.st
Weekly newsletter and directory of self-hosted apps.
Self-HPO: Self-Evolving Coding Agents via Harness-Policy Optimization
A coding agent that alternately improves its own harness and its weights (RL) on Qwen3.5-35B-A3B; the two gains compound (SWE-Verified 61.0 to 69.5). Weights are open under Apache-2.0.
Pydantic AI
Type-safe agent framework from the Pydantic team; structured outputs and dependency injection without much magic.
Open WebUI
Self-hosted chat interface for Ollama and OpenAI-compatible APIs, with RAG and multi-user support.
Ollama
The easiest way to pull and run open models locally with a simple API.
nanoGPT
The simplest readable codebase for training GPT-style models. Excellent for understanding the fundamentals.
Model Context Protocol
Specification and docs for MCP, the open standard for connecting LLM apps to tools and data.
MCP reference servers
Reference Model Context Protocol servers (filesystem, git, fetch, memory…). Useful as templates for your own tools.
llm (CLI)
Simon Willison's CLI and Python library for working with many LLMs, with logging to SQLite.
llama.cpp
LLM inference in C/C++ on CPUs and consumer GPUs; the backbone of most local model tooling and the GGUF format.
LangGraph
Stateful, graph-based agent orchestration with checkpoints and human-in-the-loop. Good when an agent workflow needs to be explicit and resumable.
Hugging Face Learn
Free courses on LLMs, agents, MCP and reinforcement learning.
Hermes Agent
Nous Research's self-improving agent with memory and skills, built to run on open models.
GrosserNL/workers-sdk (fork)
My fork of Cloudflare's Wrangler / Workers SDK, where I prepare upstream fixes.
GrosserNL/react-remove-scroll-bar
My modernised fork of react-remove-scroll-bar: zero dependencies, SSR-safe, RTL and CSP-nonce support, with tests on React 18 and 19.
GrosserNL/hermes-agent (fork)
My working fork of Nous Research's Hermes Agent, used for experiments and upstream contributions.
Gitea
Lightweight self-hosted Git service; I mirror my repositories to it.
DSPy
Program LLM pipelines instead of hand-tuning prompts, and optimise them against a metric.
CrewAI
Role-based multi-agent framework; quick to prototype teams of specialised agents.
CoworkEnv: Synthesizing Training Environments for Long-Horizon Cowork Agents
A pipeline that generates executable, grader-checked workplace tasks for SFT and RL; SFT lifts an 8B Qwen3-VL model from 45.9 to 77.8 on Claw-Eval multi-turn tasks.
Coolify
Self-hostable alternative to Heroku/Vercel for deploying apps and databases on your own servers.
Building effective agents (Anthropic)
Clear, practical guide to agent patterns: start with simple workflows and add autonomy only where it pays off.
awesome-selfhosted
The big curated list of free, self-hostable software.
AutoGen
Microsoft's framework for multi-agent conversations and event-driven agent systems.