LLM Integration

Overview

Empire Chain provides seamless integration with various Large Language Models (LLMs). This guide covers how to use different LLMs, configure them, and build applications with them.

Supported Models

OpenAI GPT Models
Anthropic Claude
Local Models (via HuggingFace)
Custom Model Integration

Basic Usage

Setting Up

from empire_chain.docling import LLMHandler

# Initialize with OpenAI
llm = LLMHandler(provider="openai")

# Initialize with Anthropic
llm = LLMHandler(provider="anthropic")

# Initialize with local model
llm = LLMHandler(
    provider="local",
    model_path="path/to/model"
)

Simple Queries

# Basic completion
response = llm.complete("Tell me about AI")

# Chat completion
messages = [
    {"role": "user", "content": "What is machine learning?"}
]
response = llm.chat(messages)

Advanced Features

Model Configuration

config = {
    "model": "gpt-4",
    "temperature": 0.7,
    "max_tokens": 150,
    "top_p": 1,
    "frequency_penalty": 0,
    "presence_penalty": 0
}

llm = LLMHandler(provider="openai", config=config)

Streaming Responses

for chunk in llm.stream("Tell me a story"):
    print(chunk, end="", flush=True)

Function Calling

functions = [{
    "name": "get_weather",
    "description": "Get weather information",
    "parameters": {
        "type": "object",
        "properties": {
            "location": {"type": "string"},
            "unit": {"type": "string"}
        }
    }
}]

response = llm.chat(
    messages=[{"role": "user", "content": "What's the weather in London?"}],
    functions=functions
)

Integration Patterns

RAG Implementation

from empire_chain.docling import RAGSystem

# Initialize RAG with specific LLM
rag = RAGSystem(llm_handler=llm)

# Add documents
rag.add_documents(["doc1.pdf", "doc2.pdf"])

# Query with context
response = rag.query("What do the documents say about AI?")

Chain of Thought

prompt = """
Question: {question}
Let's approach this step by step:
1) First, let's understand what we're asked
2) Then, break down the problem
3) Finally, provide the solution
"""

response = llm.complete(
    prompt.format(question="How does photosynthesis work?"),
    temperature=0.3
)

Agent Implementation

from empire_chain.docling import Agent

# Create an agent with tools
agent = Agent(
    llm_handler=llm,
    tools=[
        "calculator",
        "web_search",
        "code_executor"
    ]
)

# Run agent
result = agent.run("Calculate the compound interest on $1000")

Best Practices

Error Handling

from empire_chain.exceptions import LLMError

try:
    response = llm.complete("Generate text")
except LLMError as e:
    print(f"LLM Error: {e}")
except Exception as e:
    print(f"Unexpected error: {e}")

Rate Limiting

llm = LLMHandler(
    provider="openai",
    config={
        "rate_limit": {
            "requests_per_minute": 60,
            "tokens_per_minute": 40000
        }
    }
)

Caching

llm = LLMHandler(
    provider="openai",
    config={
        "cache": {
            "enabled": True,
            "ttl": 3600,  # 1 hour
            "max_size": 1000
        }
    }
)

Model Comparison

Provider	Strengths	Use Cases
OpenAI	State-of-the-art performance	General purpose, code generation
Anthropic	Long context, reasoning	Document analysis, complex tasks
Local	Privacy, no latency	Edge deployment, offline use

Security Considerations

API Key Management

# Use environment variables
import os
llm = LLMHandler(
    provider="openai",
    api_key=os.getenv("OPENAI_API_KEY")
)

Content Filtering ```python llm = LLMHandler( provider="openai", config={ "content_filter": { "enabled": True, "level": "strict" } } )