mod.json

{
    "geode": "5.7.1",
    "gd": {
        "win": "2.2081",
        "android": "2.2081",
        "mac": "2.2081",
        "ios": "2.2081"
    },
    "id": "entity12208.edit-ai",
    "name": "Editor AI",
    "version": "v2.1.9",
    "developer": "entity12208",
    "description": "Generate Geometry Dash levels using AI!",
    "dependencies": {
        "geode.node-ids": ">=v1.0.0"
    },
    "incompatibilities": {
        "alk.editor-collab": {
            "breaking": false,
            "version": "*"
        }
    },
    "settings": {
        "_general-header": {
            "type": "title",
            "name": "General"
        },
        "ai-provider": {
            "type": "string",
            "default": "gemini",
            "name": "AI Provider",
            "description": "Which backend to call. Configure the matching section below.",
            "one-of": [
                "gemini",
                "claude",
                "openai",
                "openrouter",
                "ministral",
                "huggingface",
                "deepseek",
                "ollama",
                "lm-studio",
                "llama-cpp",
                "custom",
                "manual"
            ]
        },
        "difficulty": {
            "type": "string",
            "default": "medium",
            "name": "Difficulty",
            "description": "Target difficulty for the AI. Canned values: easy / medium / hard / extreme. Free-form text is also accepted (e.g. 'demon', 'tutorial')."
        },
        "style": {
            "type": "string",
            "default": "modern",
            "name": "Style",
            "description": "Visual / gameplay style. Canned values: modern / retro / flow / memory. Free-form text is also accepted."
        },
        "length": {
            "type": "string",
            "default": "medium",
            "name": "Length",
            "description": "Approximate level length. Canned values: short / medium / long / xl / xxl. Free-form text is also accepted."
        },
        "max-objects": {
            "type": "int",
            "default": 500,
            "min": 10,
            "max": 1000000,
            "name": "Max Objects",
            "description": "Upper bound on objects the AI may emit per generation."
        },
        "target-object-count": {
            "type": "int",
            "default": 0,
            "min": 0,
            "max": 20000,
            "name": "Target Objects",
            "description": "Minimum total objects a generation must reach. The AI is asked to continue until it gets there (or extension rounds run out). <cg>0 = off</c>. Capped by Max Objects."
        },
        "edit-target-ops": {
            "type": "int",
            "default": 1500,
            "min": 0,
            "max": 5000,
            "name": "Edit Workload Target",
            "description": "When the AI reworks an existing level, it must total at least this many edits (objects moved + deleted + added + restyled) before the result stages — it is asked to continue until it gets there. Small follow-up tweaks (under 30 edits) stage immediately. <cg>0 = off</c>."
        },
        "spawn-batch-size": {
            "type": "int",
            "default": 8,
            "min": 1,
            "max": 100,
            "name": "Spawn Speed",
            "description": "Objects placed per editor tick during preview. Higher = faster but jankier."
        },
        "example-level-ids": {
            "type": "string",
            "default": "",
            "name": "Example Level IDs",
            "description": "Up to 5 comma-separated GD level IDs the AI downloads as style references at the start of every generation (requires AI tools). Leave empty for none."
        },
        "ai-ground-y": {
            "type": "int",
            "default": 105,
            "min": 15,
            "max": 300,
            "name": "Ground Y",
            "description": "Editor Y of the visible ground line. 105 is default. Lower if blocks float; higher if they bury."
        },
        "_provider-header": {
            "type": "title",
            "name": "Provider — Gemini"
        },
        "gemini-model": {
            "type": "string",
            "default": "gemini-3-flash",
            "name": "Gemini Model",
            "description": "Use <cg>Fetch model list</c> in the in-mod Provider tab to pull the live list, or type any Gemini model id."
        },
        "gemini-api-key": {
            "type": "string",
            "default": "",
            "name": "Gemini API Key",
            "description": "From Google AI Studio. Stays on this device in Geode save data."
        },
        "_provider-claude": {
            "type": "title",
            "name": "Provider — Claude"
        },
        "claude-model": {
            "type": "string",
            "default": "claude-sonnet-4-6",
            "name": "Claude Model",
            "description": "Use <cg>Fetch model list</c> in the in-mod Provider tab to pull the live list, or type any Claude model id."
        },
        "claude-api-key": {
            "type": "string",
            "default": "",
            "name": "Claude API Key",
            "description": "From console.anthropic.com. Stays on this device."
        },
        "_provider-openai": {
            "type": "title",
            "name": "Provider — OpenAI"
        },
        "openai-model": {
            "type": "string",
            "default": "gpt-4o",
            "name": "OpenAI Model",
            "description": "Use <cg>Fetch model list</c> in the in-mod Provider tab to pull the live list, or type any OpenAI model id."
        },
        "openai-api-key": {
            "type": "string",
            "default": "",
            "name": "OpenAI API Key",
            "description": "From platform.openai.com. Stays on this device."
        },
        "_provider-mistral": {
            "type": "title",
            "name": "Provider — Mistral"
        },
        "ministral-model": {
            "type": "string",
            "default": "ministral-8b-latest",
            "name": "Mistral Model",
            "description": "Use <cg>Fetch model list</c> in the in-mod Provider tab to pull the live list, or type any Mistral model id."
        },
        "ministral-api-key": {
            "type": "string",
            "default": "",
            "name": "Mistral API Key",
            "description": "From console.mistral.ai. Stays on this device."
        },
        "_provider-hf": {
            "type": "title",
            "name": "Provider — Hugging Face"
        },
        "huggingface-model": {
            "type": "string",
            "default": "meta-llama/Llama-3.1-8B-Instruct",
            "name": "HF Model",
            "description": "Repo-style id, e.g. meta-llama/Llama-3.1-8B-Instruct."
        },
        "huggingface-api-key": {
            "type": "string",
            "default": "",
            "name": "HF Token",
            "description": "From huggingface.co/settings/tokens. Stays on this device."
        },
        "_provider-openrouter": {
            "type": "title",
            "name": "Provider — OpenRouter"
        },
        "openrouter-model": {
            "type": "string",
            "default": "google/gemini-2.5-flash",
            "name": "OpenRouter Model",
            "description": "Provider/model slug, e.g. google/gemini-2.5-flash or anthropic/claude-3.5-sonnet."
        },
        "openrouter-api-key": {
            "type": "string",
            "default": "",
            "name": "OpenRouter API Key",
            "description": "From openrouter.ai/keys. One key, many hosted models."
        },
        "_provider-deepseek": {
            "type": "title",
            "name": "Provider — DeepSeek"
        },
        "deepseek-model": {
            "type": "string",
            "default": "deepseek-chat",
            "name": "DeepSeek Model",
            "description": "Use <cg>Fetch model list</c> in the in-mod Provider tab to pull the live list, or type any DeepSeek model id."
        },
        "deepseek-api-key": {
            "type": "string",
            "default": "",
            "name": "DeepSeek API Key",
            "description": "From platform.deepseek.com."
        },
        "_provider-ollama": {
            "type": "title",
            "name": "Provider — Ollama"
        },
        "use-platinum": {
            "type": "bool",
            "default": false,
            "name": "Use Platinum",
            "description": "Route Ollama requests through the EditorAI Platinum cloud proxy instead of localhost."
        },
        "ollama-model": {
            "type": "string",
            "default": "entity12208/editorai:deepseek",
            "name": "Ollama Model Tag",
            "description": "Tag as shown by `ollama list`. e.g. entity12208/editorai:v3-7b."
        },
        "ollama-timeout": {
            "type": "int",
            "default": 600,
            "min": 60,
            "max": 1800,
            "name": "Ollama Timeout (s)",
            "description": "How long to wait for a local Ollama response before giving up."
        },
        "_provider-lmstudio": {
            "type": "title",
            "name": "Provider — LM Studio"
        },
        "lm-studio-url": {
            "type": "string",
            "default": "http://localhost:1234",
            "name": "LM Studio URL"
        },
        "lm-studio-model": {
            "type": "string",
            "default": "default",
            "name": "LM Studio Model",
            "description": "Loaded model name in LM Studio. Start the server before generating."
        },
        "_provider-llamacpp": {
            "type": "title",
            "name": "Provider — llama.cpp"
        },
        "llama-cpp-url": {
            "type": "string",
            "default": "http://localhost:8080",
            "name": "llama.cpp URL"
        },
        "llama-cpp-model": {
            "type": "string",
            "default": "default",
            "name": "llama.cpp Model",
            "description": "Model name (often 'default' unless your server expects something specific)."
        },
        "_provider-custom": {
            "type": "title",
            "name": "Provider — Custom (BYOPAK)"
        },
        "custom-provider-name": {
            "type": "string",
            "default": "Custom Provider",
            "name": "Display Name",
            "description": "Cosmetic only. Shown in the AI popup info panel."
        },
        "custom-provider-url": {
            "type": "string",
            "default": "",
            "name": "Endpoint URL",
            "description": "Base URL or full /chat/completions URL. OpenAI-compatible bodies are used."
        },
        "custom-provider-api-key": {
            "type": "string",
            "default": "",
            "name": "API Key",
            "description": "Leave blank if your server needs no auth."
        },
        "custom-provider-model": {
            "type": "string",
            "default": "",
            "name": "Model",
            "description": "Model name expected by your endpoint."
        },
        "custom-provider-auth": {
            "type": "string",
            "default": "Authorization: Bearer ${KEY}",
            "name": "Auth Header Template",
            "description": "${KEY} is replaced with your API key. Clear to send no header."
        },
        "_bypass-header": {
            "type": "title",
            "name": "Bypass"
        },
        "bypass-char-filter": {
            "type": "bool",
            "default": false,
            "name": "Character Filter Bypass",
            "description": "Let every text box in the game accept any printable character (same behavior as Eclipse Menu's bypass). Off = vanilla filtering."
        },
        "bypass-char-limit": {
            "type": "bool",
            "default": false,
            "name": "Character Limit Bypass",
            "description": "Remove the maximum text length on every text box in the game (same behavior as Eclipse Menu's bypass). Off = vanilla limits."
        },
        "ui-anim-speed": {
            "name": "UI Animation Speed",
            "description": "How fast the EditorAI panel opens and closes. <cg>1</c> = slow fade (2 seconds), <cg>10</c> = instant.",
            "type": "int",
            "default": 8,
            "min": 1,
            "max": 10
        },
        "_advanced-header": {
            "type": "title",
            "name": "Advanced"
        },
        "enable-ai-tools": {
            "type": "bool",
            "default": true,
            "name": "Enable AI Tools",
            "description": "Let the AI call helper tools (web search, level analysis, song lookup, etc.) before composing."
        },
        "enable-advanced-features": {
            "type": "bool",
            "default": true,
            "name": "Triggers & Colors",
            "description": "Allow the AI to emit color/move/pulse triggers and group IDs."
        },
        "enable-rate-limiting": {
            "type": "bool",
            "default": true,
            "name": "Rate Limiting",
            "description": "Throttle generations. Leave on for hosted APIs to avoid quota burns."
        },
        "rate-limit-seconds": {
            "type": "int",
            "default": 3,
            "min": 1,
            "max": 60,
            "name": "Rate Limit (s)",
            "description": "Minimum seconds between generations when rate limiting is on."
        },
        "show-ai-output": {
            "type": "bool",
            "default": false,
            "name": "Save AI Output",
            "description": "Write each generation's full raw AI response to last-ai-response.txt in this mod's save folder. Off by default."
        },
        "enable-rating": {
            "type": "bool",
            "default": true,
            "name": "Rate Generations",
            "description": "Show the thumbs up/down rating prompt after generation."
        },
        "max-feedback-examples": {
            "type": "int",
            "default": 3,
            "min": 1,
            "max": 10,
            "name": "Feedback Examples",
            "description": "How many past rated generations to use as few-shot context."
        },
        "refinement-rounds": {
            "type": "int",
            "default": 3,
            "min": 0,
            "max": 10,
            "name": "Refinement Rounds",
            "description": "How many self-review passes the AI runs after the initial draft. 0 disables refinement."
        },
        "enable-self-critique": {
            "type": "bool",
            "default": true,
            "name": "Final Self-Check",
            "description": "One extra round after refinements where the AI scores its own level and patches the weakest area (or confirms it's good). Costs at most one additional API call."
        },
        "enable-vision": {
            "name": "Vision (level snapshots)",
            "description": "When the model supports images (Claude, Gemini, GPT-4o, LLaVA...), attach a rendered snapshot of the level to review and follow-up turns so the AI can SEE what it built.",
            "type": "bool",
            "default": true
        },
        "two-pass-generation": {
            "type": "bool",
            "default": true,
            "name": "Two-Pass Generation",
            "description": "Generate the gameplay skeleton first; after you accept it, a second pass adds decoration, color triggers, and atmosphere without touching gameplay."
        },
        "subagent-provider": {
            "type": "string",
            "default": "",
            "name": "Subagent Provider",
            "description": "Optional second model the main AI can consult via the ask_subagent tool (e.g. a big cloud model asking your local Ollama for object suggestions, or vice versa). Empty disables the tool.",
            "one-of": [
                "",
                "gemini",
                "claude",
                "openai",
                "openrouter",
                "ministral",
                "huggingface",
                "deepseek",
                "ollama",
                "lm-studio",
                "llama-cpp"
            ]
        },
        "subagent-model": {
            "type": "string",
            "default": "",
            "name": "Subagent Model",
            "description": "Model name for the subagent provider. Empty uses that provider's configured model."
        },
        "copilot-mode": {
            "type": "bool",
            "default": false,
            "name": "Copilot Mode",
            "description": "While you edit, EditorAI watches for death zones and physics problems and automatically proposes a fix blueprint (you still accept or deny). At most one auto-fix every 2 minutes."
        },
        "allow-telemetry": {
            "type": "bool",
            "default": false,
            "name": "Auto-Share Generations (Telemetry)",
            "description": "Opt-in. While ON, <cy>every generation</c> is uploaded automatically to the community training collector: prompt, settings, generated objects, and ratings (yours and the AI's self-review). Never your identity, never API keys. Levels rated above 5 become training data for the free community models."
        }
    },
    "resources": {
        "files": [
            "resources/object_ids.json"
        ]
    },
    "tags": [
        "editor",
        "utility",
        "content"
    ],
    "links": {
        "source": "https://github.com/Entity12208/EditorAI",
        "community": "https://discord.gg/5hwCqMUYNj",
        "homepage": "https://editorai.pages.dev"
    },
    "about": "Generate Geometry Dash levels by describing them in plain text.\n\n<cy>How to use</c>\nPress <cg>E</c> anywhere (rebindable in Settings; on mobile, tap the floating <cg>AI</c> bubble) to open the EditorAI panel.\n- <cy>Chat</c>: hit <cg>+ new chat</c>, pick a target - the current editor, a brand-new level, or any local level - describe what you want, and Generate. Each generation is a conversation: watch it think and work, then keep talking to it for as long as you like - small edits, big reworks, plans, questions. Conversations survive restarts and pick up where they left off.\n- <cy>Settings</c>: choose among 12 AI providers (Gemini, Claude, OpenAI, OpenRouter, Mistral, HuggingFace, DeepSeek, Ollama/Platinum, LM Studio, llama.cpp, custom, and Manual - copy-paste with any chatbot, no key), paste your API key, and tune generation.\n\nGenerated objects appear on their own editor layer - the rest of the level fades back - and you accept, edit, or deny the result. Edits the AI makes to existing objects are previewed the same way and fully reversible until you accept."
}