fogsight

Purpose

Fogsight (Mist Elephant) is an animation engine agent driven by a large language model (LLM). Its goal is to transform abstract concepts or words into high-level, vivid animations. Users only need to input a topic and click generate to obtain a high-level animation with a complete narrative, including bilingual narration and cinematic visual quality.

Main Capabilities

• Concept as imagery: Input a topic, and Fogsight will generate a high-level animation with a complete narrative for you, including bilingual narration and cinematic visual quality.
• Intelligent orchestration: At its core is LLM-driven orchestration capability. From narration and visual elements to dynamic effects, AI will automatically complete the entire creative process.
• Language User Interface (LUI): Through multiple rounds of dialogue with AI, users can precisely tune and iterate the animation until the desired artistic effect is achieved.

Usage

Online Use

You can directly visit the website fogsight.ai to use it, generating animations online without deployment.

Local Deployment

1. Environment requirements: a modern web browser (such as Chrome, Firefox, Edge); an API key for a large language model. Officially, only Google Gemini 2.5 is recommended.
2. Clone the code repository:

git clone https://github.com/fogsightai/fogsight.git
cd fogsight

1. Install dependencies:

pip install -r requirements.txt

1. Configure the API key:

cp demo-credentials.json credentials.json

Edit the credentials.json file and fill in your API_KEY and BASE_URL. Please note that the project uses an OpenAI-compatible SDK, but you should still use Gemini 2.5 pro.

1. One-click start:

python start_fogsight.py

Run the start_fogsight.py script. It will automatically start the backend service and automatically open in the browser http://127.0.0.1:8000。

1. Start creating: Enter a topic on the page (for example, "bubble sort"), then wait for the result to be generated.

Using Docker

1. Make sure Docker and docker-compose are installed on the system.
2. Configure the API key:

cp demo-credentials.json credentials.json

Edit the credentials.json file and fill in your API_KEY, BASE_URL, and MODEL.

Configuration example using OpenRouter:

{
  "API_KEY": "sk-or-v1-您的OpenRouter密钥",
  "BASE_URL": "https://openrouter.ai/api/v1",
  "MODEL": "anthropic/claude-sonnet-4"
}

Configuration example using Gemini:

{
  "API_KEY": "您的Gemini密钥",
  "BASE_URL": "",
  "MODEL": "gemini-2.5-pro"
}

1. Start using Docker Compose:

# 使用默认端口 8000
docker-compose up -d
# 或者指定自定义端口（例如 3000）
HOST_PORT=3000 docker-compose up -d

If the Docker image cannot be pulled, you can try using a proxy, or use a domestic mirror source for the image.

1. Access the application: Open a browser and visit http://localhost:8000（或您指定的端口）。
2. Stop the service:

docker-compose down

Applicable Scenarios

• Transform abstract concepts or words into vivid animations, for example, input topics such as "bubble sort" to generate narrative animations.
• Animation creation that requires bilingual narration and cinematic visual quality.
• Creative workflows that hope to precisely tune and iterate animations through multiple rounds of dialogue.

Limitations

• Local deployment requires an API key for a large language model. Officially, only Google Gemini 2.5 is recommended.
• The project is released under the CC BY-NC-ND 4.0 license, prohibiting commercial use and modification/derivatives. If you plan to use it in a commercial environment, please contact the project team.
• Information regarding fees, pricing, etc. is not yet provided.