Quick Start Guide

This guide will get you up and running with CellSem LLM Client in minutes.

Basic Usage

Creating Agents

The simplest way to get started is using the configuration utilities:

from cellsem_llm_client import create_openai_agent, create_anthropic_agent

# Create agents with automatic environment configuration
openai_agent = create_openai_agent()
anthropic_agent = create_anthropic_agent()

Simple Queries

# Basic text generation
openai_response = openai_agent.query("Explain quantum computing in 50 words")
print(openai_response)

claude_response = anthropic_agent.query("What are the benefits of renewable energy?")
print(claude_response)

Custom Configuration

# Custom model and parameters
custom_agent = create_openai_agent(
    model="gpt-4",
    max_tokens=2000
)

response = custom_agent.query("Write a detailed analysis of machine learning")

Working with System Messages

System messages help set the context and behavior of the LLM:

# Using system messages for better control
agent = create_openai_agent()

response = agent.query(
    "What is 2+2?",
    system_message="You are a math tutor. Always show your work and explain step by step."
)

Provider-Specific Features

OpenAI Models

from cellsem_llm_client import OpenAIAgent

# Using specific OpenAI models
gpt4_agent = OpenAIAgent(
    model="gpt-4",
    api_key="your-api-key",
    max_tokens=1500
)

# GPT-3.5 for faster, cheaper responses
gpt35_agent = OpenAIAgent(
    model="gpt-3.5-turbo",
    api_key="your-api-key"
)

Anthropic Models

from cellsem_llm_client import AnthropicAgent

# Using different Claude models
haiku_agent = AnthropicAgent(
    model="claude-3-haiku-20240307",  # Fast and economical
    api_key="your-api-key"
)

sonnet_agent = AnthropicAgent(
    model="claude-3-sonnet-20240229",  # Balanced performance
    api_key="your-api-key"
)

Advanced Usage

Direct LiteLLM Integration

For maximum flexibility, use the base LiteLLMAgent:

from cellsem_llm_client import LiteLLMAgent

# Works with any LiteLLM-supported provider
agent = LiteLLMAgent(
    model="gpt-4",  # or "claude-3-sonnet-20240229", etc.
    api_key="your-api-key",
    max_tokens=1000
)

response = agent.query("Your prompt here")

Environment-Based Configuration

The client automatically loads configuration from environment variables:

import os
from cellsem_llm_client import create_litellm_agent

# Set environment variables
os.environ["OPENAI_API_KEY"] = "your-openai-key"

# Agent automatically uses the environment key
agent = create_litellm_agent(model="gpt-3.5-turbo")

Best Practices

Error Handling

from cellsem_llm_client import create_openai_agent

try:
    agent = create_openai_agent()
    response = agent.query("Your prompt")
    print(response)
except ValueError as e:
    print(f"Configuration error: {e}")
except Exception as e:
    print(f"API error: {e}")

Cost Optimization

Tip

Use GPT-3.5-turbo or Claude Haiku for simple tasks
Use GPT-4 or Claude Sonnet for complex reasoning
Set appropriate max_tokens to control costs
Consider caching responses for repeated queries

Cost Tracking

Track token usage and costs for better budget management:

from cellsem_llm_client.agents.agent_connection import OpenAIAgent
from cellsem_llm_client.tracking.cost_calculator import FallbackCostCalculator

# Set up cost calculation
calculator = FallbackCostCalculator()
calculator.load_default_rates()

agent = OpenAIAgent(model="gpt-3.5-turbo", api_key="your_key")

# Query with cost tracking
response, usage = agent.query_with_tracking(
    "Explain machine learning briefly",
    cost_calculator=calculator
)

print(f"Input tokens: {usage.input_tokens}")
print(f"Output tokens: {usage.output_tokens}")
print(f"Estimated cost: ${usage.estimated_cost_usd:.6f}")
print(f"Provider: {usage.provider}")

Warning

API Key Sharing: The ApiCostTracker returns aggregate usage for the entire API key, including other applications. For precise per-request tracking, rely on the estimated costs from query_with_tracking() instead.

Testing and Development

# Use different models for development vs production
import os

if os.getenv("ENVIRONMENT") == "production":
    agent = create_openai_agent(model="gpt-4")
else:
    agent = create_openai_agent(model="gpt-3.5-turbo")  # Cheaper for testing

Example Applications

Content Generation

agent = create_anthropic_agent()

blog_post = agent.query(
    "Write a 300-word blog post about sustainable technology",
    system_message="You are a technical writer specializing in environmental technology."
)

Code Analysis

agent = create_openai_agent(model="gpt-4")

code_review = agent.query(
    f"Review this Python code for best practices:\n\n{code}",
    system_message="You are a senior Python developer. Provide constructive feedback."
)

Data Analysis

agent = create_anthropic_agent()

analysis = agent.query(
    f"Analyze this dataset summary: {data_summary}",
    system_message="You are a data scientist. Provide insights and recommendations."
)

Next Steps

Explore the cellsem_llm_client for detailed API documentation
Check out Contributing for development guidelines
See ROADMAP.md for upcoming features:
- Token tracking and cost monitoring
- JSON schema validation
- File attachment support
- AI-powered model recommendations