> ## Knowledge Base Index
> Fetch the complete knowledge base index at: https://documentation.celestory.io/sitemap.xml
> Use this file to discover available pages before exploring further.
> Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash).

# Creation Assistance & Vibe Coding with AI

# 🤖 Creation Assistance & Vibe Coding with AI
Celestory completely rethinks the relationship between the logic designer and artificial intelligence. AI is no longer treated as a decorative option bolted on at the end of a project, but as a genuine co-pilot riding along the whole way, from design to the player's experience. 🚀
This synergy rests on two pillars: creation assistance (Vibe Coding) for the designer 👨‍🎨, and interactive in-game AI for the end user 🎮.
---
## 1. ✨ What Is "Vibe Coding" in Celestory?
"Vibe Coding" is a modern design philosophy where the creator steers the development environment no longer by laboriously hunting through menus or writing a complex grammar, but by expressing themselves naturally, through intent (the vibe). 🌊
In Celestory Creator, this takes the shape of a suite of interconnected tools able to turn a simple intention (text) into instant structure or visuals.
### A. 🏗️ Structural Generation (Workflows & Logic)
Thanks to the language models built into the Celestory cloud, instead of placing twenty blocks by hand to "create a generative dialogue system wired to a database", the AI can build and wire the fundamental blocks for you. 🪄
- Living templates: you save hours of architecture work, with the AI handling the technical plumbing (making sure the JSON coming out of the LLM matches what your document base expects).
- The embedded IDE: for the rare spots that need actual code (complex JSON object handling, custom CSS, an HTML5 page), Celestory opens the Monaco editor — the one from Visual Studio Code — right inside the block edit window: syntax highlighting, line numbers, auto-closing brackets. 💻
### B. 🎭 Prototyping and Mock-ups (Dummy Data)
Building an application with no content is hard. During the wireframe phase:
- The creator can use the embedded AI to generate ultra-realistic test data (placeholder text & images) in a few words ("Generate 5 fantasy character sheets in JSON", or "Generate 3 painted forest background images"). 🖼️
- The module becomes immediately playable and testable on screen, long before your final artists or writers deliver the definitive assets.
---
## 2. 🎨 Embedded AI for the Creator: Managing Your Assets
The **AI Generation** window (`Cmd+Shift+G`) is the designer's art studio: four tabs — **Text, Image, Video, Audio** — plus a history of your **Generations**. Under the hood, text goes through **OpenRouter** (which gives access to many models: Mistral, OpenAI, Anthropic…) and media through **Fal**, known for its very low latency. 🖌️
### ⚡ Fast Visual Generation
- No more endless digging through stock image libraries. Write a text prompt ("Cyberpunk hero, neon, realistic 8k") and the image lands straight in your project. 🏎️💨
🔺 Generations consume **AI Credits**: the window shows your usage for the month, and a ⭐ marks the blocks that consume credits in the list.
- That flexibility keeps the creator in the flow (the state of concentration), with no need to leave the software to go looking for the right puzzle piece. The image ends up synchronised automatically into the project's Google Cloud Storage. ☁️
---
## 3. 🎮 In-Game AI: A Dynamic Experience for the End User
Beyond behind-the-scenes help for the no-code developer, AI becomes a major actor in the experience your user will live through (player, colleague, learner). 🌟
### A. 💬 The AI Request Blocks (Text and Generators)
There are specialised native blocks (such as the **AI text request** block), each with its own input and output points. These blocks turn Celestory into a powerful LLM (Large Language Model) API orchestrator.
- Dynamic Scenario: instead of writing out every branch of a story (**Choice** blocks), send the user's free-form answer to an LLM with hidden instructions (the system prompt: "You are a suspicious guard, you must decide whether to let the player through based on their excuse."). The LLM analyses, and Celestory steers the graph according to its yes or no. 🕵️‍♂️
- Structured Formatting: with no-code support, Celestory can require an LLM to answer exclusively in JSON. That JSON structures a piece of data (e.g. {mood: "sad", damage_taken: 20}) which then feeds into the application's numeric variables. 📊
### B. 🌅 On-the-Fly Imaging Blocks (In-Game Generation)
The tool does not stop at text. Multimodal generation blocks offer a fascinating ability: changing the world of the game or the application while the user is using it. ✨
- **AI image request**: a player describes a whimsical spell or the look of their creature. The block catches the user's text, sends it to the image AI, and two seconds later Celestory applies the generated image as an avatar or a background! 🧙‍♂️
- The Magic of Immediacy: this kind of request runs server-side, for complete API key security and minimal latency, and it revolutionises classic web interfaces by making them 100% reactive to the audience's imagination. 🔮
---
## 4. 🤖 The Assistant That Builds the Graph
"Vibe Coding" does not stop at content generation: from the **Text** tab of the AI Generation window, you can **ask the assistant to build your graph**.
- Describe what you want ("add a welcome dialogue that asks for the first name, then two choices"), and the assistant creates the blocks, wires them together, names the variables and writes the dialogue lines.
- **Actions are shown, not hidden**: every tool it calls is displayed in the conversation, and the blocks appear before your eyes on the canvas.
- A discreet button under your message lets you **put the project back as it was before that turn**, if you do not like the result.
🔺 **Not every model can do this.** If the selected model does not support tools, a banner warns you: it will answer in the conversation without ever touching the graph. Pick a tool-capable model if you want the assistant to actually build.
🔺 The assistant works on the project **open in your editor**. Remember to save (`Cmd+S`) whatever matters before launching an ambitious turn.
---
## Conclusion: 🤝 The Maker + Machine Team
In Celestory, artificial intelligence is no longer an optional final API call: it is the mortar that dramatically speeds up visual composition at the start of a project (Vibe Coding), guarantees the designer a constant flow of inspiration, and offers the module's end consumer literally infinite replayability and personalisation. 🚀
