Providers & Models
PageSpace supports 7 AI providers with 100+ models through a unified interface built on the Vercel AI SDK.
Provider Overview
PageSpace (Default)
The built-in provider uses OpenRouter with PageSpace's API key. No configuration required.
- Free tier: Includes models like Qwen3, DeepSeek R1, Mistral Small
- No API key needed: Works out of the box
- Rate limited: 50 AI calls/day on free plan
OpenRouter
Access 200+ models from multiple providers through a single API key.
Settings > AI > OpenRouter > Enter API key
Models include Claude, GPT, Gemini, Llama, Mixtral, and many more. OpenRouter handles routing, rate limiting, and fallbacks.
Google AI
Direct access to Google's Gemini models.
Settings > AI > Google AI > Enter API key
| Model | Context | Vision | Tools |
|---|---|---|---|
| Gemini 2.5 Pro | 1M tokens | Yes | Yes |
| Gemini 2.5 Flash | 1M tokens | Yes | Yes |
| Gemini 2.0 Flash | 1M tokens | Yes | Yes |
OpenAI
Direct access to OpenAI models.
Settings > AI > OpenAI > Enter API key
| Model | Context | Vision | Tools |
|---|---|---|---|
| GPT-5 | 128K tokens | Yes | Yes |
| GPT-4.1 | 1M tokens | Yes | Yes |
| GPT-4.1 mini | 1M tokens | Yes | Yes |
| o3 | 200K tokens | Yes | Yes |
| o4-mini | 200K tokens | Yes | Yes |
Anthropic
Direct access to Anthropic's Claude models.
Settings > AI > Anthropic > Enter API key
| Model | Context | Vision | Tools |
|---|---|---|---|
| Claude Opus 4.6 | 200K tokens | Yes | Yes |
| Claude Sonnet 4.6 | 200K tokens | Yes | Yes |
| Claude Haiku 4.5 | 200K tokens | Yes | Yes |
xAI
Direct access to xAI's Grok models.
Settings > AI > xAI > Enter API key
| Model | Context | Vision | Tools |
|---|---|---|---|
| Grok 4 | 256K tokens | Yes | Yes |
Ollama (Local)
Run AI models locally on your own hardware. Requires a running Ollama instance.
Settings > AI > Ollama > Configure base URL (default: http://localhost:11434)
Any model available in your Ollama installation is automatically discovered. Popular choices: Llama 3, CodeLlama, Mistral, Phi-3.
Configuration Hierarchy
AI settings can be configured at three levels, with the most specific level winning:
- User default — Your global AI provider/model preference
- Drive override — Different provider/model for a specific workspace
- Page override — Different provider/model for a specific AI chat page
// User default
Settings > AI > Default Provider: "anthropic"
Settings > AI > Default Model: "claude-sonnet-4-20250514"
// Drive override
Drive Settings > AI > Provider: "google"
Drive Settings > AI > Model: "gemini-2.5-pro-preview"
// Page override (AI_CHAT pages only)
Page Settings > AI Provider: "openai"
Page Settings > AI Model: "gpt-4.1"
API Key Security
API keys are stored encrypted in the user_ai_settings table:
- One key per provider per user
- Keys are encrypted at rest using server-side encryption
- Keys are never exposed in API responses
- Keys can be updated or deleted from Settings > AI
Model Capability Detection
PageSpace automatically detects whether models support:
- Vision (image processing): Checked via static lookup tables and pattern matching on model names
- Tool calling (function execution): Checked via OpenRouter's capability API or pattern-based fallback
When a model doesn't support tools, PageSpace:
- Falls back to text-only mode
- Informs the user
- Suggests tool-capable alternatives from the same provider
Free Models
The PageSpace provider includes several free models via OpenRouter:
- Qwen3 Coder — Good for code generation
- DeepSeek R1 — Strong reasoning model
- Mistral Small — Fast general-purpose model
Free models have usage limits based on your subscription plan. All free models support tool calling.