Grok Provider

The Grok provider implements support for xAI's Grok models through their API.

Key Features

• Full model support: Access to all available Grok models
• OpenAI-compatible API: Familiar interface for existing users
• Async-first implementation: Optimized for concurrent operations
• Streaming support: Real-time response handling
• Usage tracking: Detailed metrics and cost estimation

Supported Models

from aicore.models_metadata import METADATA

# List available Grok models
grok_models = [model for model in METADATA if model.startswith("grok-")]
print(grok_models)

Configuration

Python Configuration

from aicore.llm.config import LlmConfig

config = LlmConfig(
    provider="grok",
    api_key="your_xai_api_key",
    model="grok-3",
    temperature=0.7,
    max_tokens=1000
)

YAML Configuration

provider: grok
api_key: "your_xai_api_key"
model: "grok-3"
temperature: 0.7
max_tokens: 1000

Usage Examples

Basic Completion

from aicore.llm import Llm

llm = Llm(config=config)
response = llm.complete("Explain quantum computing in simple terms")
print(response)

Usage Tracking

usage = llm.get_usage()
print(f"Tokens used: {usage.total_tokens}")
print(f"Estimated cost: ${usage.estimated_cost:.4f}")

Requirements

• xAI API key (available through xAI developer program)

Limitations

• Currently only supports text completion (no chat interface)
• Limited model availability compared to other providers