# AI Models
ToolFront's [Python SDK](./python_sdk.md) supports various AI model providers through [Pydantic AI](https://ai.pydantic.dev/models/overview/).
[OpenAI](https://ai.pydantic.dev/models/openai/), [Anthropic](https://ai.pydantic.dev/models/anthropic/), [Google](https://ai.pydantic.dev/models/gemini/), [Bedrock](https://ai.pydantic.dev/models/bedrock/), [Cohere](https://ai.pydantic.dev/models/cohere/), [Groq](https://ai.pydantic.dev/models/groq/), [Mistral](https://ai.pydantic.dev/models/mistral/), [HuggingFace](https://ai.pydantic.dev/models/openai/#hugging-face), [xAI](https://ai.pydantic.dev/models/openai/#xai), [DeepSeek](https://ai.pydantic.dev/models/openai/#deepseek), [Azure AI](https://ai.pydantic.dev/models/openai/#azure), [Fireworks AI](https://ai.pydantic.dev/models/openai/#fireworks), [Moonshot AI](https://ai.pydantic.dev/models/openai/#moonshot), [GitHub Models](https://ai.pydantic.dev/models/openai/#github-models), [Heroku AI](https://ai.pydantic.dev/models/openai/#heroku), [OpenRouter](https://ai.pydantic.dev/models/openai/#openrouter), [Ollama](https://ai.pydantic.dev/models/openai/#ollama), [Together AI](https://ai.pydantic.dev/models/openai/#together-ai)
---
## Basic Usage
Set your model provider's API key as an environment variable, and then specify your Browser's model using the `provider:model-name` format.
=== ":simple-openai:{ .middle } OpenAI"
```bash
export OPENAI_API_KEY="your-api-key"
```
Pass the model to the browser:
```python hl_lines="3"
from toolfront import Browser
browser = Browser(model="openai:gpt-5")
result = browser.ask("Your question", url="file:///path/to/toolsite")
```
=== ":simple-anthropic:{ .middle } Anthropic"
```bash
export ANTHROPIC_API_KEY="your-api-key"
```
Pass the model to the browser:
```python hl_lines="3"
from toolfront import Browser
browser = Browser(model="anthropic:claude-sonnet-4-5")
result = browser.ask("Your question", url="file:///path/to/toolsite")
```
=== ":simple-google:{ .middle } Google"
```bash
export GOOGLE_API_KEY="your-api-key"
```
Pass the model to the browser:
```python hl_lines="3"
from toolfront import Browser
browser = Browser(model="google-gla:gemini-2.5-pro")
result = browser.ask("Your question", url="file:///path/to/toolsite")
```
=== ":simple-mistralai:{ .middle } Mistral"
```bash
export MISTRAL_API_KEY="your-api-key"
```
Pass the model to the browser:
```python hl_lines="3"
from toolfront import Browser
browser = Browser(model="mistral:mistral-large-latest")
result = browser.ask("Your question", url="file:///path/to/toolsite")
```
=== ":simple-huggingface:{ .middle } HuggingFace"
```bash
export HUGGINGFACE_API_KEY="your-api-key"
```
Pass the model to the browser:
```python hl_lines="3"
from toolfront import Browser
browser = Browser(model="huggingface:Qwen/Qwen3-235B-A22B")
result = browser.ask("Your question", url="file:///path/to/toolsite")
```
!!! tip "Default Model"
Set a default model with the `TOOLFRONT_MODEL` environment variable:
```bash
export TOOLFRONT_MODEL="openai:gpt-5"
```
---
## Custom Providers
You can use any [Pydantic AI model](https://ai.pydantic.dev/models/overview/) directly for custom or local models.
```python
from toolfront import Browser
from pydantic_ai.models.openai import OpenAIChatModel
ollama_model = OpenAIChatModel(
'llama3.2',
base_url='http://localhost:11434/v1',
api_key='ignored',
)
browser = Browser(model=ollama_model)
result = browser.ask("Your question", url="file:///path/to/toolsite")
```
