Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Houdini Agent - DCC Asset Manager with AI capabilities
| Date | Stars |
|---|---|
| 2026-07-31 | 330 |
| 2026-08-06 | 333 |
Today
+3 stars today
This week
— stars this week
This month
— stars this month
Momentum
47.0
growth rate 0.00%/day
# Houdini Agent **[English](README.md)** | **[中文](README_CN.md)** A standalone AI assistant for SideFX Houdini, featuring autonomous multi-turn tool calling, **Meshy AI 3D generation** (text/image→3D, concept galleries, image-to-image editing, retexture, remesh), a **cloud asset library**, web search, VEX/Python code execution, Plan mode for complex tasks, a brain-inspired long-term memory system, a plugin hook system for community extensions, user-defined context rules, and a modern **QML/Qt Quick** desktop UI with bilingual support. As of **v2.0**, Houdini Agent runs as a **standalone desktop app** that connects to a running Houdini over a lightweight **bridge** (no longer embedded in the Houdini process), with a completely rewritten **QML/Qt Quick** front-end (Mono Editorial design). Built on the **OpenAI Function Calling** protocol, the agent can read node networks, create/modify/connect nodes, run VEX wrangles, generate and import 3D assets via Meshy, execute system shell commands, search the web, query local documentation, create structured execution plans, learn from past interactions, and be extended via plugins — all within an iterative agent loop. A centralized **ToolRegistry** unifies core tools, skills, and plugin tools with mode-based access control. ## Core Features ### Agent Loop The AI operates in an autonomous **agent loop**: it receives a user request, plans the steps, calls tools, inspects results, and iterates until the task is complete. Three modes are available: - **Agent mode** — Full access to all 50+ tools. The AI can create, modify, connect, and delete nodes, set parameters, execute scripts, generate and import 3D assets, and save the scene. - **Ask mode** — Read-only. The AI can only query scene structure, inspect parameters, search documentation, and provide analysis. All mutating tools are blocked by a `ToolRegistry` mode guard. - **Plan mode** — The AI enters a planning phase: it researches the current scene (read-only), clarifies requirements via `ask_question`, then generates a structured execution plan with DAG flow diagram. The user reviews and confirms before execution begins. An **auto-resume mechanism** detects premature AI termination and forces continuation until all steps are complete. ``` User request → AI plans → call tools → inspect results → call more tools → … → final reply ``` - **Multi-turn tool calling** — the AI decides which tools to call and in what order - **Todo task system** — complex tasks are broken into tracked subtasks with live status updates - **Streaming output** — real-time display of thinking process and responses - **Extended Thinking** — native support for reasoning models (DeepSeek-R1, GLM-4.7, Claude with `<think>` tags) - **Stop anytime** — interrupt the running agent loop at any point - **Smart context management** — round-based conversation trimming that never truncates user/assistant messages, only compresses tool results - **Long-term memory** — brain-inspired three-layer memory system (episodic, semantic, procedural) with reward-driven learning and automatic reflection - **Plugin system** — external community extensions via `plugins/` directory with hook events, custom tools, UI buttons, and settings - **User Rules** — Cursor-style persistent context rules that are automatically injected into every AI request ### Supported AI Providers | Provider | Models | Notes | |----------|--------|-------| | **DeepSeek** | `deepseek-v4-flash`, `deepseek-v4-pro`, `deepseek-chat`*, `deepseek-reasoner`* | V4: explicit thinking param + reasoning_effort; *old models deprecated 2026/07/24 | | **GLM (Zhipu AI)** | `glm-4.7` | Stable in China, native reasoning & tool calling | | **OpenAI** | `gpt-5.2`, `gpt-5.3-codex` | Powerful, full Function Calling & Vision support | | **Duojie** (relay) | `claude-opus-4-6-gemini`, `claude-opus-4-6-max`, `claude-sonnet-4-5`, `claude-sonnet-4-6`, `gemini-3-flash`, `gemini-3.1-pro`, `glm-5-turbo`, `glm-5.1`, `MiniMax-M2.7`, `Min
Excerpt of 67,771 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:9df4d29399612b66, llm:Repository title and description: 'Houdini Agent - DCC Asset Manager with AI capabilities' (DCC = Digital Content Creation). Suggests an asset management tool for graphics/animation with AI features, likely integrates agents for asset workflows in 3D content creation.
matched fp:9df4d29399612b66, llm:Repository title and description: 'Houdini Agent - DCC Asset Manager with AI capabilities' (DCC = Digital Content Creation). Suggests an asset management tool for graphics/animation with AI features, likely integrates agents for asset workflows in 3D content creation.
matched fp:9df4d29399612b66, llm:Repository title and description: 'Houdini Agent - DCC Asset Manager with AI capabilities' (DCC = Digital Content Creation). Suggests an asset management tool for graphics/animation with AI features, likely integrates agents for asset workflows in 3D content creation.
matched fp:9df4d29399612b66, llm:Repository title and description: 'Houdini Agent - DCC Asset Manager with AI capabilities' (DCC = Digital Content Creation). Suggests an asset management tool for graphics/animation with AI features, likely integrates agents for asset workflows in 3D content creation.