Configuration

Planova's Settings page lets you configure language, pipeline behavior, and AI provider connections. Open it from the top navigation bar.

Language

Planova supports two interface languages:

Code	Language
en-US	English
zh-CN	Chinese (Simplified)

Click the language button to switch. The change takes effect immediately across the entire UI.

Pipeline Mode

Planova offers two pipeline modes for parsing floor plans:

Hybrid CV+VLM (default)

Uses classical computer vision (edge detection, contour analysis) for wall and room geometry, combined with a Vision Language Model for semantic understanding (room labels, furniture, door/window types). This mode generally produces more accurate wall geometry.

Legacy

The VLM handles everything -- both geometry extraction and semantic analysis. This mode is simpler but may produce less precise wall positions on complex floor plans.

:::tip When to Use Legacy Mode Try Legacy mode if the Hybrid mode produces strange geometry for your particular floor plan style. Some hand-drawn or non-standard floor plans work better with the VLM-only approach. :::

LLM Provider Configuration

Planova uses three separate AI providers, each serving a different purpose in the pipeline. You configure each one independently.

All three providers share the same configuration fields:

Field	Description	Example
Base URL	The API endpoint URL	https://api.openai.com/v1
API Key	Your authentication key	sk-...
Model	The model identifier to use	gpt-4o

Each provider card has a Test Connection button that sends a lightweight request to verify your credentials and measure latency.

VLM Provider (Image Parsing)

This provider runs the Vision Language Model that reads your floor plan image and extracts structural data: rooms, walls, doors, windows, and their positions.

Requirements: The model must support image input (multimodal). Examples:

• gpt-4o (OpenAI)
• claude-3.5-sonnet (Anthropic)
• qwen-vl-max (Alibaba)

Chat Provider (Furniture Planning)

This provider runs the language model that decides what furniture to place in each room, based on the room type, size, and your chosen style preset.

Requirements: Any capable chat/completion model. Examples:

• gpt-4o (OpenAI)
• claude-3.5-sonnet (Anthropic)
• deepseek-chat (DeepSeek)

Image Provider (AI Rendering)

This provider generates photorealistic renders of your 3D scene from a given camera angle and prompt.

Requirements: An image generation API. Examples:

• dall-e-3 (OpenAI)
• A self-hosted Stable Diffusion endpoint

:::warning Separate Providers Are Fine The three providers do not need to come from the same vendor. You can use OpenAI for VLM, DeepSeek for chat, and a local Stable Diffusion instance for rendering -- whatever gives you the best results and cost for each task. :::

Testing Your Connection

For each provider, click Test Connection after filling in the Base URL, API Key, and Model fields. The test will:

• Validate that the endpoint is reachable
• Verify your API key is accepted
• Confirm the model exists
• Report the round-trip latency in milliseconds

A green checkmark with latency (e.g. "Connection successful (342ms)") means everything is working. A red indicator with an error message will guide you toward the issue.

Data Storage Location

Planova stores all data in the platform-specific application data directory:

Platform	Path
Linux	~/.local/share/com.planova.app/
macOS	~/Library/Application Support/com.planova.app/
Windows	C:\Users\<you>\AppData\Roaming\com.planova.app\

This directory contains:

• planova.db -- SQLite database with projects, files, and settings
• uploads/ -- original uploaded floor plan images
• previews/ -- generated preview images
• pipeline/ -- intermediate pipeline artifacts (per project)

:::tip Backing Up Your Data To back up all your projects, copy the entire com.planova.app directory. The SQLite database is self-contained and portable. :::

Saving Settings

After making changes, click the Save button at the top of the Settings page. Settings are persisted to the SQLite database and take effect immediately.

:::info Settings Are Per-Installation All settings (including API keys) are stored locally on your machine. They are never sent to any external service except the AI providers you explicitly configure. :::