camel-py 0.2.0

A lightweight Python client for Ollama with context management and streaming support.

🐪 Camel-py

A lightweight Python client for Ollama with built-in agentic capabilities and tool calling support.

---

🚀 Installation

pip install camel-py

⚡ Quickstart

Basic Chat

from camel import CamelClient

with CamelClient(model="llama3") as client:
    resp = client.chat("Hello, who are you?")
    print(resp.text)
    
    # Streaming
    print("Assistant: ", end="")
    client.stream("Tell me a joke about camels")

AI Agent with Tool Calling

from camel import CamelClient, Agent, Tool

def get_weather(location: str) -> str:
    return f"Weather in {location}: 72°F, sunny"

weather_tool = Tool(
    name="get_weather",
    description="Gets current weather for any location",
    schema={
        "type": "object",
        "properties": {
            "location": {"type": "string", "description": "City name"}
        },
        "required": ["location"]
    },
    function=get_weather
)

client = CamelClient(model="llama3")
agent = Agent(client, tools=[weather_tool])

result = agent.run("What's the weather in Paris?")
print(result)

🔧 Features

• AI Agents: Built-in tool calling via FunctionGemma (auto-installed)
• Dual-model architecture: Specialized tool detection + your choice for responses
• Streaming: Real-time token streaming
• Context management: Save/load/clear conversation history
• Model management: List, pull, delete Ollama models
• Embeddings: Generate text embeddings

📂 Examples

• examples/agent.py → AI agent with tool calling
• examples/adv_chat.py → Context persistence

🛠️ How It Works

The Agent uses a two-model approach:

1. FunctionGemma detects when tools are needed
2. Your chosen model generates natural responses

This provides reliable tool calling while maintaining conversation quality.

📦 Requirements

• Python ≥3.12
• Ollama running locally