Tag
Explore by tags
Detects, segments, and tracks every instance of an open-vocabulary concept in images and video from a text phrase or visual exemplar, not just one object per prompt. An 848M-param model reaching ~75-80% of human accuracy across 270K concepts.
Unifies agentic tasks, reasoning, and coding in a single MoE model with 355B total / 32B active parameters and a switchable thinking mode. A lighter 106B-param Air variant trades scale for efficiency; both ship MIT-licensed.
Closes a learning loop most agents lack: turns experience into reusable skills, refines them mid-task, and full-text searches its own past sessions for recall. Runs from CLI or Telegram/Discord/Slack and schedules unattended cron jobs.
Modular marketplace of focused Claude Code plugins that composes specialized agents and progressive 'agent skills' to orchestrate multi-agent development workflows while minimizing token usage.
Coordinates specialized AI agents — developer, browser, document, multimodal — running in parallel on your desktop to automate multi-step work. Runs fully local via Ollama, vLLM, or LM Studio, with built-in MCP tools and human-in-the-loop checkpoints.
Parses PDF resumes into structured JSON using LLMs, enriches profiles with GitHub signals, and outputs explainable category scores, evidence, bonuses and deductions. Runs fully local with Ollama or via Google Gemini; designed for reproducible, fairness-constrained resume scoring in hiring workflows.
Compiles an agent's raw chat logs, documents, and tool traces into three persistent layers — index, learned skills, and user memory — so context survives sessions. Claims 92% Locomo-benchmark accuracy and up to 95% lower token cost than replaying history.
Collaborates on web tasks in real time: edit its plan before it runs, pause and grab the browser mid-task, and approve irreversible clicks before they happen. A research prototype for studying human-in-the-loop oversight instead of full autonomy.
Provides a long‑lived, in‑process file and content search library for editors and AI agents, with typo‑resistant fuzzy matching, frecency‑ranked results, background watchers, and a lightweight in‑memory content index — optimized for repeated searches in long‑running processes.
Provides ultra-fast, typo-tolerant file search and grep tuned for Neovim and AI agents, with built-in memory (frecency, git status, size, definition matches). It reduces agent token use and speeds developer file discovery in large repos.
Indexes any repo into a knowledge graph of dependencies, call chains, and execution flows, then feeds it to AI coding agents via MCP so they stop missing context. Ships as a CLI plus a zero-install browser graph explorer with chat.
A 20B-parameter MMDiT diffusion model that generates and edits images with accurate embedded text, including dense Chinese and English typography. Handles complex multi-line layouts and identity-preserving edits while keeping text legible.