import asyncio
from SimpleLLMFunc import llm_function, OpenAICompatible

# Load LLM interface from configuration file
llm = OpenAICompatible.load_from_json_file("provider.json")["your_provider"]["model"]

@llm_function(llm_interface=llm)
async def classify_sentiment(text: str) -> str:
    """
    Analyze the sentiment tendency of text.

    Args:
        text: Text to analyze

    Returns:
        Sentiment classification, can be 'positive', 'negative', or 'neutral'
    """
    pass  # Prompt as Code!

async def main():
    result = await classify_sentiment("This product is amazing!")
    print(f"Sentiment classification: {result}")

asyncio.run(main())

"""
Example using LLM function decorator
"""
import asyncio
from typing import List
from pydantic import BaseModel, Field
from SimpleLLMFunc import llm_function, OpenAICompatible, app_log

# Define a Pydantic model as return type
class ProductReview(BaseModel):
    rating: int = Field(..., description="Product rating, 1-5 points")
    pros: List[str] = Field(..., description="List of product advantages")
    cons: List[str] = Field(..., description="List of product disadvantages")
    summary: str = Field(..., description="Review summary")

# Use decorator to create an LLM function
@llm_function(
    llm_interface=OpenAICompatible.load_from_json_file("provider.json")["volc_engine"]["deepseek-v3-250324"]
)
async def analyze_product_review(product_name: str, review_text: str) -> ProductReview:
    """You are a professional product review expert who needs to objectively analyze the following product review and generate a structured review report.
    
    The report should include:
    1. Overall product rating (1-5 points)
    2. List of main product advantages
    3. List of main product disadvantages
    4. Summary evaluation
    
    Rating rules:
    - 5 points: Perfect, almost no disadvantages
    - 4 points: Excellent, advantages clearly outweigh disadvantages
    - 3 points: Average, advantages and disadvantages are basically equal
    - 2 points: Poor, disadvantages clearly outweigh advantages
    - 1 point: Very poor, almost no advantages
    
    Args:
        product_name: Name of the product to review
        review_text: User's review content of the product
        
    Returns:
        A structured ProductReview object containing rating, advantages list, disadvantages list, and summary
    """
    pass  # Prompt as Code, Code as Doc

async def main():
    
    app_log("Starting example code")
    # Test product review analysis
    product_name = "XYZ Wireless Headphones"
    review_text = """
    I've been using these XYZ wireless headphones for a month. The sound quality is very good, especially the bass performance is excellent,
    and they're comfortable to wear, can be used for long periods without fatigue. The battery life is also strong, can last about 8 hours after full charge.
    However, the connection is occasionally unstable, sometimes suddenly disconnects. Also, the touch controls are not sensitive enough, often need to click multiple times to respond.
    Overall, these headphones have great value for money, suitable for daily use, but if you need them for professional audio work, they might not be enough.
    """
    
    try:
        print("\n===== Product Review Analysis =====")
        result = await analyze_product_review(product_name, review_text)
        # result is directly a Pydantic model instance
        # no need to deserialize
        print(f"Rating: {result.rating}/5")
        print("Advantages:")
        for pro in result.pros:
            print(f"- {pro}")
        print("Disadvantages:")
        for con in result.cons:
            print(f"- {con}")
        print(f"Summary: {result.summary}")
    except Exception as e:
        print(f"Product review analysis failed: {e}")

if __name__ == "__main__":
    asyncio.run(main())

from SimpleLLMFunc import llm_chat, tui


@tui(custom_event_hook=[...])
@llm_chat(llm_interface=my_llm_interface, stream=True, enable_event=True)
async def agent(message: str, history=None):
    """Your agent prompt"""


if __name__ == "__main__":
    agent()

from SimpleLLMFunc.hooks import is_event_yield

async for output in agent("split this into parallel subtasks"):
    if not is_event_yield(output):
        continue

    if output.origin.fork_id:
        print(
            f"[fork:{output.origin.fork_id} depth={output.origin.fork_depth}] "
            f"{output.event.event_type}"
        )
    else:
        print(f"[main] {output.event.event_type}")

from SimpleLLMFunc import llm_function, llm_chat


@llm_function(llm_interface=my_llm_interface)
async def async_analyze_text(text: str) -> str:
    """Async text content analysis"""
    pass


@llm_chat(llm_interface=my_llm_interface, stream=True)
async def async_chat(message: str, history: List[Dict[str, str]]):
    """Async chat functionality, supports streaming responses"""
    pass


async def main():
    result = await async_analyze_text("Text to analyze")

    async for response, updated_history in async_chat("Hello", []):
        print(response)

@llm_function(llm_interface=my_llm_interface, max_tool_calls=None)
async def analyze(text: str) -> str:
    """Analyze the text."""
    pass


@llm_chat(llm_interface=my_llm_interface, stream=True, max_tool_calls=12)
async def cautious_agent(message: str, history=None):
    """Chat agent with an explicit safety cap."""
    pass

from SimpleLLMFunc import llm_function
from SimpleLLMFunc.type import ImgPath, ImgUrl, Text

@llm_function(llm_interface=my_llm_interface)
async def analyze_image(
    description: Text,           # Text description
    web_image: ImgUrl,          # Web image URL
    local_image: ImgPath        # Local image path
) -> str:
    """Analyze images and provide detailed explanations based on descriptions
    
    Args:
        description: Specific requirements for image analysis
        web_image: Web image URL to analyze
        local_image: Local reference image path for comparison
        
    Returns:
        Detailed image analysis results
    """
    pass

import asyncio


async def main():
    result = await analyze_image(
        description=Text("Please describe the differences between these two images in detail"),
        web_image=ImgUrl("https://example.com/image.jpg"),
        local_image=ImgPath("./reference.jpg")
    )
    print(result)


asyncio.run(main())

@llm_function(
    llm_interface=llm_interface,          # LLM interface instance
    toolkit=[tool1, tool2],                # Tool list
    retry_on_exception=True,               # Auto retry on exception
    timeout=60                              # Timeout setting
)
async def my_function(param: str) -> str:
    """Supports {language} {style} analysis"""
    pass

result = await my_function(
    "some input",
    _template_params={
        "language": "English",
        "style": "Professional",
    },
)

from SimpleLLMFunc import OpenAICompatible

# Method 1: Load from JSON configuration file
provider_config = OpenAICompatible.load_from_json_file("provider.json")
llm = provider_config["deepseek"]["v3-turbo"]

# Method 2: Direct creation
llm = OpenAICompatible(
    api_key="sk-xxx",
    base_url="https://api.deepseek.com/v1",
    model="deepseek-chat"
)

@llm_function(llm_interface=llm)
async def my_function(text: str) -> str:
    """Process text"""
    pass

from SimpleLLMFunc.interface import LLM_Interface

class CustomLLMInterface(LLM_Interface):
    async def call_llm(self, messages, **kwargs):
        # Implement your own LLM calling logic
        pass

from SimpleLLMFunc.logger import app_log, push_error, log_context

# 1. Basic logging
app_log("Starting request processing", trace_id="request_123")
push_error("Error occurred", trace_id="request_123", exc_info=True)

# 2. Use context manager to automatically associate logs
with log_context(trace_id="task_456", function_name="analyze_text"):
    app_log("Starting text analysis")  # Automatically inherits context trace_id
    try:
        # Execute operations...
        app_log("Analysis completed")
    except Exception:
        push_error("Analysis failed", exc_info=True)  # Also automatically inherits trace_id

from pydantic import BaseModel, Field
from SimpleLLMFunc.tool import tool

# Define Pydantic model for complex parameters
class Location(BaseModel):
    latitude: float = Field(..., description="Latitude")
    longitude: float = Field(..., description="Longitude")

# Use decorator to create tool
@tool(name="get_weather", description="Get weather information for specified location")
async def get_weather(location: Location, days: int = 1) -> dict:
    """
    Get weather forecast for specified location
    
    Args:
        location: Location information, including latitude and longitude
        days: Forecast days, default is 1 day
        
    Returns:
        Weather forecast information
    """
    # Actual implementation would call weather API
    return {
        "location": f"{location.latitude},{location.longitude}",
        "forecast": [{"day": i, "temp": 25, "condition": "Sunny"} for i in range(days)]
    }

from SimpleLLMFunc.tool import tool
from SimpleLLMFunc.type import ImgPath, ImgUrl

@tool(name="generate_chart", description="Generate charts based on data")
async def generate_chart(data: str, chart_type: str = "bar") -> ImgPath:
    """
    Generate charts based on provided data
    
    Args:
        data: CSV format data
        chart_type: Chart type, default is bar chart
        
    Returns:
        Generated chart file path
    """
    # Actual implementation would generate chart and save locally
    chart_path = "./generated_chart.png"
    # ... Chart generation logic
    return ImgPath(chart_path)

@tool(name="search_web_image", description="Search web images")
async def search_web_image(query: str) -> ImgUrl:
    """
    Search web images
    
    Args:
        query: Search keywords
        
    Returns:
        Found image URL
    """
    # Actual implementation would call image search API
    image_url = "https://example.com/search_result.jpg"
    return ImgUrl(image_url)