Beyond basic AI conversations, Cloosphere Chat offers three advanced features: web search, image generation, and code execution.
Click the ”+” button on the left of the input box to open the capability toggles and file attachment menu.
Activation Conditions
Advanced features only appear in the menu when all of the following are satisfied.
| Condition | Description |
|---|
| System setting | Admin must enable the feature globally |
| User permission | Permissions like permissions.features.web_search are required |
| Model capability | The agent’s capability setting must not be off |
Admins (admin role) bypass permission checks, so as long as system settings are enabled, they can use the feature.
Feature Details
Web Search
Image Generation
Code Interpreter
Web Search
The AI searches the web in real time and answers based on the latest information.
How to Enable
Open the toggle menu
Click the ”+” button on the left of the input box.
Enable Web Search
Toggle on Web Search. After enabling, web search is applied to subsequent messages.
How It Works
- The AI auto-extracts a search query from the user’s question
- Performs web search via the admin-configured search engine (SearxNG, Google PSE, Brave, etc.)
- Search result URL list and query are shown collapsed
- Generates an answer using the search results as context
- Provides source links as citations
Search Result Display
The search query and referenced URL list appear at the top of the response in collapsible form.
Click each URL to navigate to the original page.In Settings > Interface > Web Search in Chat, set web search to “always enabled” to auto-search in every conversation without toggling.
Use Cases
- “What’s today’s KOSPI index?”
- “What are the latest AI trends?”
- “Find recent news about this company”
Image Generation
Use image generation models like DALL-E to generate images from text prompts.
How to Enable
When multiple image generation Connections are configured, choose which Connection to use from the submenu.Open the toggle menu
Click the ”+” button on the left of the input box.
Pick Image Generation
Click Image Generation (may appear as Image depending on admin settings) — available Connections appear.
Pick a Connection
Pick the image generation engine to use. Hover over each Connection to see details like model name, size, and quality.
Connection Info
Each image Connection includes:| Item | Description |
|---|
| Deployment | Image generation deployment name (Azure deployment name) |
| Model | Model name |
| Size | Output image size (e.g., 1024x1024) |
| Quality | Image quality setting (when applicable) |
| Format | Output format (PNG, JPEG, etc.) |
Prompt Tips
- Be specific: Instead of “mountain”, say “snow-covered Alps at sunrise, photorealistic”
- Specify style: “watercolor”, “minimal illustration”, “3D render”
- State composition: “wide shot”, “close-up”, “aerial view”
For agent models, admins can restrict to specific Connections. In that case, only allowed Connections appear in the list.
Code Interpreter
Run AI-generated code directly and see results.
How to Enable
Open the toggle menu
Click the ”+” button on the left of the input box.
Enable Code Interpreter
Toggle on Code Interpreter.
Supported Features
| Feature | Description |
|---|
| Python execution | Run Python via Pyodide (browser) or server-side Jupyter (per admin setting) |
| Data visualization | Generate charts using matplotlib, plotly, etc. |
| File generation | Download code execution outputs as files |
| Result display | Real-time stdout, stderr, and return value display |
Code Block Execution
When the AI generates a response containing code blocks, a Run button appears at the top of each block.
Clicking runs the code and displays results below the block.Artifact Viewer
When the AI generates HTML or SVG code, the Artifact Viewer panel auto-opens to display the rendered result.Request web content
Ask the AI to generate web pages, charts, SVG graphics, UI components, etc.
Auto-render
When HTML/SVG is generated, the Artifact Viewer panel auto-opens with rendered output.
Review the result
Inspect the rendered output in the Artifact panel and copy code if needed.
In browser (Pyodide) mode, system access and network requests are restricted. When admins configure server-side Jupyter, those restrictions may not apply.
| Item | Description |
|---|
| Tool name | Name of the tool registered by admin |
| Tool description | Tooltip shown on hover |
| Enable switch | Toggle Tool use in this conversation |
Tools are enabled per conversation. Enabling a Tool in one conversation doesn’t affect others.
Agent vs. Base Model
Advanced feature availability depends on the selected model type.
| Feature | Agent Model | Base Model |
|---|
| Web Search | Per Capability setting | Shown when system + user permissions are met |
| Image Generation | Per Capability setting | Shown when system + user permissions are met |
| Code Interpreter | Per Capability setting | Shown when system + user permissions are met |
In agent models, capabilities set to on are auto-enabled at conversation start.
Capabilities set to user must be toggled by the user.
