Explore the directory

An academic survey and paper collection on aligning large language models with human expectations, covering data collection, training methodologies and evaluation for alignment research.

• Papers

A series of open-weight 11-billion-parameter vision-language models for long-form and real-time video understanding, including a variant that watches a continuous video stream and decides on its own when to respond.

• Models

A toolkit for compressing large generative AI models, including language models, vision-language models and video-generation models, using quantization, pruning and token-reduction algorithms to cut size while preserving performance.

• Models
• Inference

A bridge that lets Hermes Agent, OpenClaw and OpenCode share one WeChat account at once, running two local proxy servers and a protocol bridge so each agent can be reached and switched between from the same chat.

• Agent frameworks

An agent daemon that runs AI agents, dispatches their tools and connects them to MCP servers, treating shell access, file access, skills and session search as sandboxed, independently granted capabilities rather than one open permission set.

• Agent frameworks

A set of nine local skills that give an AI coding agent a personal task-alignment system: clarifying goals before ambiguous work starts, tracking a repository-local task tree, and pausing and resuming work at safe checkpoints.

• Agent skills

A self-improving, open-source agentic harness that runs real tasks such as sending reports, managing a calendar or a CRM and booking flights, either standalone in a terminal or as the default agent inside a hosted workspace.

• Agent frameworks

An MCP server that lets a large language model manage Docker with natural language: composing containers, inspecting running ones, and managing volumes, presenting a plan for approval before applying it.

• MCP servers
• Containers

A local inference server for Apple Silicon that serves MLX models behind both OpenAI-compatible and Anthropic-compatible API endpoints, so existing SDK-based clients can point at it as a drop-in local replacement.

• Inference

A local code-intelligence MCP server and CLI that returns a symbol's context, call graph and impact analysis pre-correlated in one response, giving an AI coding agent what would otherwise take a grep-and-read loop.

• MCP servers
• Agent memory

A trace-native CI and CD tool for AI agents that turns a production failure directly into a regression test, replaying the exact recorded trace on every pull request instead of relying on hand-written evaluation datasets.

• AI evals
• CI/CD

A routing service and set of IDE plugins that sends a coding assistant's prompts to several language models at once and returns whichever answer wins on cost and quality, reached through a paid gateway API key.

• LLM gateways

The registry index that Cargo, Rust's package manager, reads to resolve crate versions and their dependencies, available both as a Git repository and as an HTTP sparse-protocol index.

• Languages

A collection of nREPL middleware, originally built for the CIDER Clojure editor integration, that adds definition lookup, code completion, and other interactive development features to any nREPL-based Clojure tool.

• Editors

A fully local, offline memory plugin for Claude Code that keeps an append-only log of each session and condenses it into a compact resume-ready summary using a classical Python summarizer, spending no model tokens.

• Agent memory

A self-paced, under-one-hour GitHub Skills exercise that walks a developer through using GitHub Copilot to explain, plan, write and review code on a sample extracurricular-activities website.

• Tutorials

A multi-machine dashboard for orchestrating many AI coding agents at once, giving each one persistent memory, a code-graph view of what it owns, and a messaging protocol so agents can coordinate directly instead of through a human relay.

• Coding agents

A self-hosted, team-oriented AI agent platform where each member gets an isolated agent workspace that reads files, runs shell commands, calls APIs and retrieves from a shared knowledge base to complete assigned tasks.

• Agent frameworks

A collection of Jupyter notebook demos for Microsoft Foundry, Azure's platform for building with models such as GPT, DALL-E, Sora, Phi, and Mistral, covering agents, image generation, embeddings, and pricing lookups.

• Tutorials

A local-first dashboard that shows what coding agents such as Claude Code, Codex and OpenCode actually did on a task, breaking each one into tool calls, files changed, tests run, time spent and estimated cost.

• AI evals

A library that adds type-safe, schema-guaranteed output to existing open language models without retraining them, letting a model keep its own free-form reasoning while its final answer stays within a declared JSON schema.

• Agent frameworks

A Python framework that connects persistent memory, model providers and reasoning so an AI agent can recall past context, reuse experience and coordinate with other agents across sessions.

• Agent memory

A production-oriented AI workflow runtime that organizes language-model calls, scripts, APIs, database actions and human approvals into versioned, retryable and auditable workflows, instead of one long-running single agent.

• Agent frameworks

An actively maintained collection of research papers on building autonomous agents, organized separately for reinforcement-learning-based agents and large-language-model-based agents.

• Papers