from llmkit import LLMKitClient, CompletionRequest, Message

# Create client from environment variables
client = LLMKitClient.from_env()

# Make a completion request
response = client.complete(CompletionRequest(
    model="anthropic/claude-sonnet-4-20250514",
    messages=[Message.user("Hello!")]
))
print(response.text_content())

from llmkit import AsyncLLMKitClient, CompletionRequest, Message

async def main():
    client = AsyncLLMKitClient.from_env()
    response = await client.complete(CompletionRequest(
        model="openai/gpt-4o",
        messages=[Message.user("Hello!")]
    ))
    print(response.text_content())

# Sync streaming
for chunk in client.complete_stream(request):
    if chunk.text:
        print(chunk.text, end="", flush=True)

# Async streaming
async for chunk in await async_client.complete_stream(request):
    if chunk.text:
        print(chunk.text, end="", flush=True)

from llmkit import ToolBuilder

# Build tools with fluent API
weather_tool = ToolBuilder("get_weather") \
    .description("Get current weather for a location") \
    .string_param("city", "City name", required=True) \
    .enum_param("unit", "Temperature unit", ["celsius", "fahrenheit"]) \
    .build()

request = CompletionRequest(
    model="anthropic/claude-sonnet-4-20250514",
    messages=[Message.user("What's the weather in Tokyo?")]
).with_tools([weather_tool])

response = client.complete(request)
for tool_call in response.tool_calls():
    print(f"Call {tool_call.name} with {tool_call.arguments}")

# Large system prompts are automatically cached
request = CompletionRequest(
    model="anthropic/claude-sonnet-4-20250514",
    messages=[
        Message.system(large_system_prompt),  # Cached after first call
        Message.user("Question 1")
    ]
).with_cache()

# Subsequent calls reuse the cached system prompt
response = client.complete(request)
print(f"Cache savings: {response.usage.cache_read_tokens} tokens")

request = CompletionRequest(
    model="anthropic/claude-sonnet-4-20250514",
    messages=[Message.user("Solve this step by step: ...")]
).with_thinking(budget_tokens=10000)

response = client.complete(request)
print("Reasoning:", response.thinking_content())
print("Answer:", response.text_content())

from llmkit import get_model_info, get_models_by_provider, get_models_with_capability

# Get model details instantly
info = get_model_info("anthropic/claude-sonnet-4-20250514")
print(f"Context: {info.context_window:,} tokens")
print(f"Input: ${info.input_price}/1M tokens")
print(f"Output: ${info.output_price}/1M tokens")

# Find models by provider
anthropic_models = get_models_by_provider("anthropic")

# Find models with specific capabilities
vision_models = get_models_with_capability(vision=True)