Skip to content

Commit 15d7b25

Browse files
entity12208entity12208
committed
Update the README and about files
1 parent bb51f6d commit 15d7b25

2 files changed

Lines changed: 98 additions & 91 deletions

File tree

‎README.md‎

Lines changed: 69 additions & 31 deletions
Original file line numberDiff line numberDiff line change
@@ -7,70 +7,108 @@ AI-powered level generation for Geometry Dash. Describe your level in plain text
77
1. Download the [latest release](https://github.com/entity12208/editorai/releases/latest)
88
2. Open the game and navigate to the Geode page
99
3. Press the manual download button and select the file
10-
4. Set up the mod using one of the AI providers below
10+
4. Open the editor and press the **AI** button
11+
5. Click the **gear icon** to configure your AI provider
1112

1213
## AI Setup
1314

15+
All settings are configured in-game via the **gear icon** in the AI popup. No need to dig through Geode mod settings.
16+
1417
### EditorAI Platinum *(Free — recommended)*
1518
[EditorAI Platinum](https://github.com/entity12208/editorai-platinum) is a community proxy that lets anyone run the official EditorAI Ollama models over the internet for free.
1619

17-
1. In mod settings, set **AI Provider** to `ollama`
18-
2. Enable the **Use Platinum** toggle
19-
3. Select a model from the **Ollama Model** dropdown:
20-
- `entity12208/editorai:qwen` — more powerful
21-
- `entity12208/editorai:deepseek` — more creative
20+
1. Open settings → set **Provider** to `ollama`
21+
2. Switch to the **Provider** tab and enable **Platinum**
22+
3. Available models are auto-detected from the Platinum network
2223

2324
Want to donate your computer to the network? Follow the instructions in the [Platinum repository](https://github.com/entity12208/editorai-platinum).
2425

2526
### Ollama *(Free — recommended for 6+ GB VRAM)*
2627
*Uses ~5.2 GB storage. Fully local and private — no internet required after setup.*
2728

2829
1. Download and install [Ollama](https://ollama.com)
29-
2. Pull the official model: `ollama pull entity12208/editorai:<model>`
30+
2. Pull the official model: `ollama pull entity12208/editorai:qwen`
3031
- `deepseek` — more creative &nbsp;|&nbsp; `qwen` — more powerful &nbsp;(each ~5.2 GB)
31-
3. In mod settings, set **AI Provider** to `ollama` and leave **Use Platinum** off
32-
4. Select your model from the **Ollama Model** dropdown
32+
3. Open settings → set **Provider** to `ollama`
33+
4. Your installed models are auto-detected on the **Provider** tab
3334

3435
### Claude *(Paid)*
3536
1. Get an API key at [console.anthropic.com](https://console.anthropic.com)
36-
2. In mod settings, set **AI Provider** to `claude` and choose a model
37-
3. Paste your key into the **Claude API Key** field in mod settings
37+
2. Open settings → set **Provider** to `claude` → switch to **Provider** tab
38+
3. Select a model and paste your API key
3839

3940
### Mistral AI *(Paid)*
4041
1. Get an API key at [console.mistral.ai](https://console.mistral.ai/api-keys/)
41-
2. In mod settings, set **AI Provider** to `ministral` and choose a model
42-
3. Paste your key into the **Mistral API Key** field in mod settings
42+
2. Open settings → set **Provider** to `ministral` → switch to **Provider** tab
43+
3. Select a model and paste your API key
4344

4445
### HuggingFace *(Free tier available)*
4546
1. Get a token at [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens)
46-
2. In mod settings, set **AI Provider** to `huggingface`
47-
3. Enter a model ID in the model field (default: `meta-llama/Llama-3.1-8B-Instruct`)
48-
4. Paste your token into the **HuggingFace API Key** field in mod settings
47+
2. Open settings → set **Provider** to `huggingface` → switch to **Provider** tab
48+
3. Enter a model ID and paste your token
4949

5050
### ChatGPT *(Paid)*
5151
1. Get an API key at [platform.openai.com/api-keys](https://platform.openai.com/api-keys)
52-
2. In mod settings, set **AI Provider** to `openai` and choose a model
53-
3. Paste your key into the **OpenAI API Key** field in mod settings
52+
2. Open settings → set **Provider** to `openai` → switch to **Provider** tab
53+
3. Select a model and paste your API key
5454

5555
### Gemini *(Not recommended — tight rate limits)*
5656
1. Get an API key at [aistudio.google.com/api-keys](https://aistudio.google.com/u/0/api-keys)
57-
2. In mod settings, set **AI Provider** to `gemini` and choose a model
58-
3. Paste your key into the **Gemini API Key** field in mod settings
57+
2. Open settings → set **Provider** to `gemini` → switch to **Provider** tab
58+
3. Select a model and paste your API key
5959

6060
## Usage
61-
1. Download and set up the mod
62-
2. Open the editor of any level
63-
3. Press the **AI** button in the top right
64-
4. Enable or disable **Clear Level**, and enter your prompt
65-
5. Click **Generate** and wait!
61+
1. Open the editor of any level
62+
2. Press the **AI** button in the editor toolbar
63+
3. Type a description of what you want (e.g. "medium difficulty platforming level")
64+
4. Click **Generate** and wait — a timer shows elapsed time
65+
5. AI-generated objects appear as **blue ghost previews**
66+
6. Choose **Accept**, **Edit**, or **Deny**:
67+
- **Accept** — makes objects permanent
68+
- **Edit** — makes objects solid so you can move/delete them, then press **Done**
69+
- **Deny** — removes all generated objects
70+
7. Optionally rate the generation 1-10 to help the AI learn your style
6671

67-
## Advanced Features
68-
Enable **Advanced Features** in mod settings to unlock:
69-
- **Color Triggers** — AI can place color triggers that smoothly transition background, ground, and object colors
70-
- **Move Triggers** — AI can create moving platforms and obstacles using grouped objects and move triggers with easing
71-
- **Group IDs** — AI assigns group IDs to objects so move triggers can target them precisely
72+
Use the **up/down arrows** next to the text input to recall previous prompts. Press **Cancel** during generation to abort.
73+
74+
## Settings
7275

73-
These features work best with smarter models (Claude, Ollama qwen, or Platinum).
76+
Click the **gear icon** in the top-left of the AI popup to access all settings, organized into three tabs:
77+
78+
| Tab | Settings |
79+
|-----|----------|
80+
| **General** | Provider, difficulty, style, length, max objects, spawn speed |
81+
| **Provider** | Model selection, API keys, Ollama/Platinum config, timeout |
82+
| **Advanced** | Triggers & colors, rate limiting, feedback & learning |
83+
84+
## Advanced Features
85+
Enable **Triggers/Colors** in the Advanced settings tab to unlock:
86+
- **Color Triggers** — smooth color transitions for background, ground, and objects
87+
- **Move Triggers** — moving platforms and obstacles with easing curves
88+
- **Rotate Triggers** — spinning decorations and obstacles around center points
89+
- **Alpha Triggers** — fade objects in and out
90+
- **Toggle Triggers** — show/hide groups of objects
91+
- **Pulse Triggers** — flash colors on groups for visual emphasis
92+
- **Spawn & Stop Triggers** — chain trigger sequences and cancel animations
93+
- **Speed Portals** — control game speed (0.5x to 4x)
94+
- **Player Visibility** — show/hide the player icon and trail
95+
- **Color Channels** — assign objects to color channels for coordinated color changes
96+
- **Group IDs** — assign objects to groups for trigger targeting
97+
- **Multi-Activate** — make orbs, pads, and triggers fire on every touch
98+
99+
These features work best with smarter models (Claude, Gemini, Ollama qwen, or Platinum).
100+
101+
## Blueprint Preview
102+
Generated objects appear as blue-tinted ghost previews (40% opacity, non-interactable) before you commit them. This lets you see what the AI made before it affects your level. The **Edit** button makes objects solid so you can rearrange or delete individual pieces before pressing **Done**.
103+
104+
## Feedback & Learning
105+
When enabled (on by default), a rating popup appears after you accept/edit/deny a generation. Your ratings and optional text feedback are stored locally and injected into future AI prompts as examples — the AI learns what you like and avoids what you don't. Feedback examples are matched by similarity to your current request (difficulty, style, length).
106+
107+
## Error Codes
108+
If something goes wrong, the error message includes a code like `[EAI-G2001]`. Share this code when reporting issues — it encodes the provider, error category, and specific problem, making diagnosis much faster.
109+
110+
## Fine-Tune Your Own Model
111+
The `training/` folder contains scripts to fine-tune a Qwen Instruct model on GD level data using QLoRA, then export it to Ollama. See [training/README.md](training/README.md) for details.
74112

75113
## Support
76114
Join the [Editor AI Discord server](https://discord.gg/5hwCqMUYNj) for support and ideas.

‎about.md‎

Lines changed: 29 additions & 60 deletions
Original file line numberDiff line numberDiff line change
@@ -7,70 +7,39 @@ AI-powered level generation for Geometry Dash. Describe your level in plain text
77
1. Download the [latest release](https://github.com/entity12208/editorai/releases/latest)
88
2. Open the game and navigate to the Geode page
99
3. Press the manual download button and select the file
10-
4. Set up the mod using one of the AI providers below
10+
4. Open the editor and press the **AI** button
11+
5. Click the **gear icon** to configure your AI provider
1112

1213
## AI Setup
1314

14-
### EditorAI Platinum *(Free — recommended)*
15-
[EditorAI Platinum](https://github.com/entity12208/editorai-platinum) is a community proxy that lets anyone run the official EditorAI Ollama models over the internet for free.
16-
17-
1. In mod settings, set **AI Provider** to `ollama`
18-
2. Enable the **Use Platinum** toggle
19-
3. Select a model from the **Ollama Model** dropdown:
20-
- `entity12208/editorai:qwen` — more powerful
21-
- `entity12208/editorai:deepseek` — more creative
22-
23-
Want to donate your computer to the network? Follow the instructions in the [Platinum repository](https://github.com/entity12208/editorai-platinum).
24-
25-
### Ollama *(Free — recommended for 6+ GB VRAM)*
26-
*Uses ~5.2 GB storage. Fully local and private — no internet required after setup.*
27-
28-
1. Download and install [Ollama](https://ollama.com)
29-
2. Pull the official model: `ollama pull entity12208/editorai:<model>`
30-
- `deepseek` — more creative &nbsp;|&nbsp; `qwen` — more powerful &nbsp;(each ~5.2 GB)
31-
3. In mod settings, set **AI Provider** to `ollama` and leave **Use Platinum** off
32-
4. Select your model from the **Ollama Model** dropdown
33-
34-
### Claude *(Paid)*
35-
1. Get an API key at [console.anthropic.com](https://console.anthropic.com)
36-
2. In mod settings, set **AI Provider** to `claude` and choose a model
37-
3. Paste your key into the **Claude API Key** field in mod settings
38-
39-
### Mistral AI *(Paid)*
40-
1. Get an API key at [console.mistral.ai](https://console.mistral.ai/api-keys/)
41-
2. In mod settings, set **AI Provider** to `ministral` and choose a model
42-
3. Paste your key into the **Mistral API Key** field in mod settings
15+
All settings are configured in-game via the **gear icon** in the AI popup.
4316

44-
### HuggingFace *(Free tier available)*
45-
1. Get a token at [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens)
46-
2. In mod settings, set **AI Provider** to `huggingface`
47-
3. Enter a model ID in the model field (default: `meta-llama/Llama-3.1-8B-Instruct`)
48-
4. Paste your token into the **HuggingFace API Key** field in mod settings
49-
50-
### ChatGPT *(Paid)*
51-
1. Get an API key at [platform.openai.com/api-keys](https://platform.openai.com/api-keys)
52-
2. In mod settings, set **AI Provider** to `openai` and choose a model
53-
3. Paste your key into the **OpenAI API Key** field in mod settings
54-
55-
### Gemini *(Not recommended — tight rate limits)*
56-
1. Get an API key at [aistudio.google.com/api-keys](https://aistudio.google.com/u/0/api-keys)
57-
2. In mod settings, set **AI Provider** to `gemini` and choose a model
58-
3. Paste your key into the **Gemini API Key** field in mod settings
59-
60-
## Usage
61-
1. Download and set up the mod
62-
2. Open the editor of any level
63-
3. Press the **AI** button in the top right
64-
4. Enable or disable **Clear Level**, and enter your prompt
65-
5. Click **Generate** and wait!
66-
67-
## Advanced Features
68-
Enable **Advanced Features** in mod settings to unlock:
69-
- **Color Triggers** — AI can place color triggers that smoothly transition background, ground, and object colors
70-
- **Move Triggers** — AI can create moving platforms and obstacles using grouped objects and move triggers with easing
71-
- **Group IDs** — AI assigns group IDs to objects so move triggers can target them precisely
72-
73-
These features work best with smarter models (Claude, Ollama qwen, or Platinum).
17+
### EditorAI Platinum *(Free — recommended)*
18+
[EditorAI Platinum](https://github.com/entity12208/editorai-platinum) is a community proxy that lets anyone use the official EditorAI models for free.
19+
20+
1. Open settings → set **Provider** to `ollama`
21+
2. Switch to the **Provider** tab and enable **Platinum**
22+
3. Models are auto-detected from the Platinum network
23+
24+
### Ollama *(Free — local)*
25+
1. Install [Ollama](https://ollama.com) and pull a model: `ollama pull entity12208/editorai:qwen`
26+
2. Open settings → set **Provider** to `ollama`
27+
3. Installed models are auto-detected
28+
29+
### Cloud Providers
30+
Claude, OpenAI, Gemini, Mistral, and HuggingFace are supported. Open settings → select your provider → switch to the Provider tab to enter your API key and select a model.
31+
32+
## Features
33+
- **Blueprint Preview** — AI objects appear as blue ghosts. Accept, edit, or deny before committing.
34+
- **Edit Mode** — make previewed objects solid, rearrange them, then press Done.
35+
- **15 Trigger Types** — color, move, rotate, alpha, toggle, pulse, spawn, stop triggers, speed portals, player visibility, and more.
36+
- **Color Channels** — assign objects to channels for coordinated color changes.
37+
- **Feedback & Learning** — rate generations 1-10 with optional text. The AI learns your style over time.
38+
- **Prompt History** — recall previous prompts with up/down arrows.
39+
- **Cancel Generation** — abort a request at any time.
40+
- **Error Codes** — every error includes a diagnostic code (e.g. `[EAI-G2001]`) for fast troubleshooting.
41+
- **In-Game Settings** — all configuration via a tabbed settings popup, no digging through Geode menus.
42+
- **Auto-Detect Models** — Ollama and Platinum models are listed automatically.
7443

7544
## Support
7645
Join the [Editor AI Discord server](https://discord.gg/5hwCqMUYNj) for support and ideas.

0 commit comments

Comments
 (0)