feat (provider/venice): Add Venice provider

[proteanx]

Venice Provider Integration

Overview

This PR adds support for Venice, a high-performance AI model provider focused on privacy and security. The integration enables both text and image generation capabilities through Venice's API.

Key Features

Text Generation

• Support for Venice's LLaMA-based models:
  • llama-3.2-3b (small but efficient)
  • llama-3.3-70b (large, powerful)
  • llama-3.1-405b (very large)
  • dolphin-2.9.2-qwen2-72b
  • qwen32b
• Optional system prompt inclusion via include_venice_system_prompt parameter
• Streaming support for real-time text generation

Image Generation

• Support for multiple image generation models:
  • fluently-xl (high-quality generation)
  • flux-dev (standard content filtering)
  • flux-dev-uncensored (no content filtering)
  • pony-realism (specialized for realism)
  • stable-diffusion-3.5
• Advanced image parameters:

  venice_parameters: {
    style_preset: "photographic", // Options: anime, photographic, digital-art
    safe_mode: true,             // Content filtering
    num_inference_steps: 30,     // Generation quality
    guidance_scale: 7.5,         // Prompt adherence
    height: 1024,               
    width: 1024,
    seed: 12345                  // Reproducibility
  }

Implementation Details

• OpenAI-compatible interface for easy integration
• Comprehensive test coverage for both text and image generation
• Environment variable support (VENICE_API_KEY)
• Custom configuration options (API endpoint, headers)
• TypeScript types and documentation

Example Usage

Text Generation:

const { text } = await generateText({
  model: venice('llama-3.3-70b'),
  prompt: 'Write a JavaScript function that sorts a list:',
});

Image Generation:

const { images } = await generateImages({
  model: venice.imageModel('fluently-xl'),
  prompt: 'A serene mountain landscape at sunset',
  venice_parameters: {
    style_preset: 'photographic',
    safe_mode: true
  }
});

Documentation

Full documentation has been added /content/providers/01-ai-sdk-providers/45-venice.mdx

Testing (can provide key if needed)

Includes test suite covering:

• Provider configuration
• Text generation
• Error handling
• Parameter validation

Dependencies

• Added @ai-sdk/venice package with minimal dependencies
• Leverages existing @ai-sdk/openai-compatible for API compatibility

[lgrammel]

Can you publish this as a 3rd party provider in a separate repository? We can then add a documentation page under community providers (please open PR for that)