Explore the directory

A launcher for llama.cpp and ik_llama.cpp that reads a GGUF model's tensor layout against the VRAM, RAM and per-GPU bandwidth actually available, then places large mixture-of-experts models across a mismatched multi-GPU setup without manual flag tuning.

• Inference
• CLI tools

A small proof-of-concept language model, explicitly not built as an LLM, that predicts in a latent space, keeps internal recurrent state, and generates output byte by byte instead of token by token.

• Models

A curated research survey covering how reinforcement learning for large language models assigns credit to individual tokens, steps, turns or agents when the only feedback is a single outcome reward at the end of a long trajectory.

• Curated lists
• Papers

A Rust-based tool layer for coding agents on OpenCode and Pi that replaces plain file reads and bash edits with symbol-aware code operations, semantic search, and background task management.

• Agent skills

A Rust library for type-safe, multi-language code generation that tracks imports, resolves naming conflicts, and produces width-aware formatted output.

• Libraries

A local-first speech-to-text engine, built in Rust on the ggml inference library, that transcribes audio entirely on-device through an OpenAI-compatible API, with no cloud calls and no telemetry.

• Speech

A reference hardware and software implementation from a research paper that gives a large language model a small physical body: an ESP32 microcontroller with sensors and outputs, plus a Node.js bridge service connecting it to the model.

• Papers

A self-hosted, browser-based workbench for building, curating, captioning and training LoRA and full image-model checkpoints, covering everything from raw image collection through fine-tuned model comparison in one interface.

• Models
• Generative media

A Kubernetes-native framework for deploying and orchestrating AI agents as cluster resources, with hierarchical multi-agent delegation and an OpenAI-compatible chat endpoint on every agent.

• Agent frameworks
• Containers

A large-language-model inference and serving engine written entirely in Java, using the TornadoVM compiler to just-in-time compile transformer kernels to CUDA, OpenCL or Apple Metal at runtime, so the same build runs on any GPU vendor.

• Inference

A curated bundle of sixteen installable skills for AI coding agents such as Claude Code, GitHub Copilot and Cursor, aimed at creative production tasks such as presentations, video editing and landing pages.

• Agent skills

GeNN (GPU-enhanced Neuronal Networks) is a code-generation-based simulation environment for spiking neural network models, compiling network descriptions to native NVIDIA CUDA or AMD HIP code for GPU execution.

• Languages

The official TypeScript and Node.js client for the Pinecone vector database, supporting full-text search, vector operations, index management, and integrated embedding and reranking.

• Databases

An agent harness for long-running coding tasks that runs Claude Code, Codex or Hermes as a multi-agent orchestrator, deciding each turn whether to spawn workers, register reusable skills, replan or stop.

• Coding agents

A local-first memory system for AI coding agents that keeps decisions, lessons and facts in a single SQLite file on the user's disk and feeds them back to Claude Code, Cursor or Codex over MCP without any cloud call.

• Agent memory

An extension for coding agents such as Pi, Claude Code and Codex that runs deterministic prose checks on a completed response and sends it back for one clearer rewrite when it crosses a configured complexity threshold.

• Agent skills

A growing series of written tutorials that build up LLM inference optimization step by step, from autoregressive generation and the KV cache through prefill-decode disaggregation to production engines such as vLLM and SGLang.

• Tutorials

A conversational digital-transformation platform, built on the RASA framework, that gives teams a no-code web interface for building, training, testing and maintaining LLM-based chatbot assistants.

• Agent frameworks
• Workflow automation

An inference toolkit for tuning and deploying deep-learning models on NVIDIA GPUs, exposing a single Python API over backends such as TensorRT, Torch-TensorRT, TorchAO, Torch Inductor and ONNX Runtime.

• Inference

A Redis-compatible database engine that also provides typed relational tables, native vector search, time series, realtime subscriptions, streams, and optional application authentication in one runtime.

• Databases

A terminal extension for the Pi coding agent that adds a status rail and a live sidebar showing the active model, context usage, Git status, cost and tool activity while a session runs.

• Coding agents

An installable Android app that packages the DeepSeek Harness AI agent runtime for a phone, letting it install apps, tap the screen, change system settings and take screenshots through Shizuku or root, without needing Termux or command-line setup.

• Agent frameworks
• Mobile

A benchmarking library for retrieval-augmented generation systems, focused on question answering, that standardizes how retrievers, rerankers and language models are compared across many datasets.

• RAG
• AI evals

A collection of runnable notebooks implementing a domain-agnostic Context Engine, a transparent architecture for orchestrating multi-agent systems through semantic context and reasoning rather than hard-coded workflow logic.

• Agent frameworks
• Tutorials