Skip to content
View Piggidragon's full-sized avatar
  • Germany

Highlights

  • Pro

Block or report Piggidragon

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Piggidragon/README.md

Hi, I'm Piggidragon πŸ‰

Selfhosting and AI nerd from Germany πŸ‡©πŸ‡ͺ
I play Minecraft, RPGs and survival games, and I build things around them.

Followers Location: Germany


πŸ”­ What I'm working on

  • πŸ§ͺ Piggidragon/llama.cpp: llama.cpp fork with quantized-native FlashAttention, prompt cache in RAM and on disk, smarter speculative decoding and multi-GPU over PCIe β€” faster long-context inference on unbalanced GPU pairs.
  • πŸ€– Pithagoras: together with Codacus I overhauled this AI agent portal almost end to end β€” the shared Documents feature, the 3D memory graph, voice mode, images, Signal as a channel, theming and a lot of the portal settings.
  • πŸ¦€ Nerve-Protocol: two Rust binaries, one MCP server. Let any agent reach any remote device with granular permission control β€” no uncontrollable SSH.
  • ⚑ optllama: a llama.cpp fork that I contribute to, focused on KV cache offload to RAM, unbalanced dual-GPU setups and partial KV cache residency.
  • 🧩 OpenWebUI-plugins: tools, functions and pipes for Open WebUI.
  • ⛏️ Minecraft modding with NeoForge: a magic and dimension mod, plus an Ender Dragon fight overhaul.

πŸ› οΈ Tech stack

Python Rust Java JavaScript LaTeX

🀝 Open source contributions

Project What I contributed
generelschwerz/llama.cpp (optllama) Host-resident KV cache: pipelined delivery, head-split cache under split mode tensor, partial KV residency budget on the slowest link first. Multi-GPU: attention split separate from the tensor split, tied output projection split, meta transport ring per device share. CUDA: quantized-native MMA FlashAttention. Also: DFlash speculative decoding, llama-bench placement and sweep flags, scheduler fixes and the arch-coverage test suite
thecodacus/pithagoras The Documents feature (shared Markdown store with git history, comment threads with @agent, word-processor editor, uploads), the 3D memory graph and Librarian chat, voice mode (orb, push-to-talk, own models), image generation and editing with gallery and viewer, the Signal channel, theming (Warm/Dusk folded into the themes extension), portal password/OIDC/updates settings, test hardening, upstream merges and release fixes β€” alongside Codacus's core work

πŸ“Š GitHub stats

GitHub stats Top languages

πŸ’¬ Find me on Discord

Discord: Piggidragon optllama Discord Codacus Discord

Ask me about selfhosting, local AI, agents and Minecraft modding.

Pinned Loading

  1. thecodacus/pithagoras thecodacus/pithagoras Public

    Pithagoras β€” hosted web UI for the pi coding agent. Give it a task, close the browser, come back when it's done.

    TypeScript 290 78

  2. Nerve-Protocol Nerve-Protocol Public

    Two Rust binaries, one MCP Server. Let any agent reach any remote device with granular permission control - No uncontrollable SSH.

    Rust 1

  3. llama.cpp llama.cpp Public

    Forked from GenerelSchwerz/llama.cpp

    LLM inference in C/C++

    C++ 1

  4. Nerve-Desktop Nerve-Desktop Public

    WIP

  5. GenerelSchwerz/llama.cpp GenerelSchwerz/llama.cpp Public

    Forked from ggml-org/llama.cpp

    LLM inference in C/C++

    C++ 144 26