Add SambaNova provider support

[Pavanmanikanta98]

Summary

Implements SambaNova Cloud provider following the same pattern as other OpenAI-compatible providers (Together, Fireworks, Alibaba).

Changes

Core Implementation

• New Provider: SambaNovaProvider class in pydantic_ai/providers/sambanova.py
  • Inherits from Provider[AsyncOpenAI]
  • Supports environment variables: SAMBANOVA_API_KEY, SAMBANOVA_BASE_URL
  • Multi-profile support for Meta Llama, DeepSeek, Qwen, and Mistral model families
  • OpenAI-compatible API endpoint: https://api.sambanova.ai/v1

Model Support

• Meta Llama: Meta-Llama-3.1-8B-Instruct, Meta-Llama-3.3-70B-Instruct, Llama-4-Maverick-17B-128E-Instruct
• DeepSeek: DeepSeek-R1-0528, DeepSeek-V3-0324, DeepSeek-R1-Distill-Llama-70B
• Qwen: Qwen3-32B
• Mistral: Mistral-based models

Files Changed

• pydantic_ai/providers/sambanova.py - New provider implementation
• pydantic_ai/providers/__init__.py - Provider registration
• pydantic_ai/models/__init__.py - Added to compatibility type aliases
• tests/providers/test_sambanova_provider.py - Comprehensive test suite (11 tests)
• tests/conftest.py - Test fixtures and httpx client cleanup
• tests/test_examples.py - Environment variable setup
• README.md - Added to provider list
• docs/index.md - Added to provider list
• docs/models/overview.md - Added to OpenAI-compatible providers
• docs/models/openai.md - Full documentation section with examples
• docs/api/providers.md - API reference

Implementation Details

Multi-Profile Pattern

The provider implements profile matching based on model name prefixes, similar to Fireworks:

prefix_to_profile = {
    'deepseek-': deepseek_model_profile,
    'meta-llama-': meta_model_profile,
    'llama-': meta_model_profile,
    'qwen': qwen_model_profile,
    'mistral': mistral_model_profile,
}

This approach matches SambaNova's actual API naming convention (direct model names without provider prefixes).

Usage Example

from pydantic_ai import Agent

# Using shorthand
agent = Agent('sambanova:Meta-Llama-3.1-8B-Instruct')

# Or explicit provider
from pydantic_ai.providers import SambaNovaProvider
provider = SambaNovaProvider(api_key='your-key')
agent = Agent(provider, model_name='DeepSeek-R1-0528')

Checklist

• Implementation follows existing provider patterns (Alibaba PR Add Alibaba Cloud DashScopeProvider and support audio input for Qwen Omni #3596, Fireworks, Together)
• Comprehensive test coverage
• Documentation updated
• All files follow project code style
• Environment variables properly configured in test fixtures

Fixes #3669

---

Note: This implementation was verified against SambaNova's actual API documentation and model catalog to ensure correct model naming conventions and profile matching.

[dsfaccini]

thanks for the contribution @Pavanmanikanta98 ! Douwe will take a look when he's back :)

[DouweM]

Note that this will have to be moved if #3941 merges first.

[Pavanmanikanta98]

@DouweM Updated with the correct link to the official SambaNova model catalog documentation.

[DouweM]

@Pavanmanikanta98 Thanks Pavan!