Explore the directory

A self-hosted AI agent workspace with a browser UI that connects to local or OpenAI-compatible models and can edit code, run shell commands, deploy Docker-backed apps, search the web, and maintain memory across sessions on your own Linux machine.

• Coding agents
• Agent frameworks

A protocol bridge that wires separate AI coding agents such as Claude Code, Codex, Cursor and Cline into one open A2A 1.0 mesh so they can discover each other, share an inbox and send each other tasks across machines over mTLS.

• Coding agents
• Agent frameworks

A QLoRA fine-tune of Qwen2.5-VL-3B that reads a driving-scene photo and outputs structured analysis of lane lines, vehicles, traffic signs and driving risk.

• Models

An out-of-tree vLLM plugin that adds Mobilint NPU support, so Mobilint-optimized language and vision-language models can be served through ordinary vLLM commands.

• Inference

A small transformer variant that merges the attention layer with the SwiGLU feed-forward layer to cut parameters at very small model sizes, without losing much performance.

• Models
• Papers

Fine-tunes IndoBART or mT5 to summarize Indonesian news articles into a headline and a short paragraph, trained on the IndoNLG and Liputan6 datasets.

• Models

An evidence-first research project benchmarking exact-attention latency and memory on consumer NVIDIA GPUs, working toward a hybrid attention operator.

• Inference

An autonomous web crawler and attack-surface-management engine for security testing, mapping endpoints, parameters and security issues across traditional and single-page web apps.

• Security

A React Native example app demonstrating NobodyWho, a library for running LLMs locally and efficiently on a phone, including chat, tool calling, vision, RAG, speech-to-text and text-to-speech.

• Inference
• Tutorials

A durable task queue implemented as versioned SQL functions inside PostgreSQL, giving TypeScript, Python and Go workers the same queue, durable waits and dashboard over one SQL protocol without a separate broker.

• Backend
• Databases

An in-memory vector database with pluggable indexing algorithms, linear, KD-tree and LSH, metadata filtering, snapshot persistence, and a FastAPI REST API plus a native Python client.

• Databases

An Apache Flink connector for the Lance columnar data format that provides source and sink support, Flink SQL and Table API integration, KNN vector search, and index building for Lance vector datasets.

• Databases
• Data engineering

A self-hosted AI agent you run on your own server and control primarily through a browser web UI, using tools such as shell access, file I/O, web search, scheduled tasks and sub-agents, and reachable from Feishu or a terminal with one shared configuration.

• Agent frameworks

A research malware triage framework built around the Model Context Protocol that combines static analysis, reverse engineering, threat-intelligence lookups, optional sandbox detonation, SHAP-based explainability and a multi-agent reporting pipeline, focused on Windows PE files.

• Security

A community-run registry for AI agent memory-system benchmark results, requiring per-question evidence records, a fixed reader, judge and prompt set per version, and reproducible harnesses instead of self-reported scores.

• AI evals

A local-first, signed communication protocol and CLI that lets AI agent sessions on different machines or under different operators exchange messages directly, with no vendor relay required unless you opt into the public wireup.net federation.

• Agent frameworks

A lightweight TypeScript vector database library that stores text chunks with metadata and embeddings in SQLite using the sqlite-vec extension, with built-in Hugging Face Transformers embedding support for cosine similarity search.

• RAG
• Databases

An example retrieval-augmented generation application that lets actuaries ask questions over life valuation actuarial documents, such as statutory accounting, principle-based reserving and IFRS 17, citing the source context behind each answer.

• RAG

A package manager for AI coding assistants that installs skills, subagents and MCP servers from any GitHub or Bitbucket repository into Claude Code, Codex, Amp, GitHub Copilot and Cursor from one command, with a security scanner and a content-hashed lockfile.

• Agent skills
• CLI tools

An early-stage toolkit for generating synthetic datasets for LLM fine-tuning, with multiple backends, filtering and GPU acceleration.

• Models

An ML-augmented successor to Snaffler that ranks files on SMB shares by likelihood of containing credentials, using a LightGBM path classifier and an optional Qwen3 content classifier to confirm flagged files.

• Security

A hierarchical multi-agent example system for AWS CloudOps that answers questions about costs, AWS Health events and network resilience through a supervisor agent that delegates to specialized FinOps, governance and operations agents.

• Agent frameworks
• Monitoring

A local-first memory library for AI agents that stores everything in one SQLite file with no required dependencies, combining exact vector and BM25 recall, conflict detection and a time-travel history, with LangChain, CrewAI and MCP integrations.

• Agent memory

The official implementation of a CVPR 2025 paper on retrieval-augmented neural fields for embedding geographic coordinates, with a pip-installable package that returns high-resolution visual-feature embeddings for any latitude and longitude.

• Models
• Papers