A Python AI agent built with LangChain that supports multiple model providers including Anthropic Claude, OpenAI GPT, and Ollama models. The agent provides a unified interface for tool calling across different model providers.

• Multi-provider support: Anthropic, OpenAI compatible APIs, and Ollama
• Unified tool interface: Same tool definitions work across all model providers
• Predefined presets: Quick setup with common model configurations
• Flexible configuration: Custom model parameters and endpoints
• Interactive CLI: Real-time conversation with tool calling capabilities

1. Install dependencies:

pip install -r requirements.txt

2. Set up API keys (as needed):

export ANTHROPIC_API_KEY="your-anthropic-key"
export OPENAI_API_KEY="your-openai-key"

The easiest way to get started is with predefined model presets:

# Anthropic Claude models
python main.py --preset claude-3.5-sonnet
python main.py --preset claude-3-haiku

# OpenAI models
python main.py --preset gpt-4-turbo
python main.py --preset gpt-3.5-turbo

# Ollama models (requires Ollama running locally)
python main.py --preset llama3.1-8b
python main.py --preset llama3.1-70b
python main.py --preset mistral-7b
python main.py --preset codellama-7b

For more control, use custom provider configurations:

# Anthropic with custom parameters
python main.py --provider anthropic --model claude-3-5-sonnet-20241022 --temperature 0.7

# OpenAI with custom API key
python main.py --provider openai --model gpt-4-turbo-preview --api-key your-key

# Custom OpenAI-compatible endpoint
python main.py --provider openai --model custom-model --base-url https://api.custom.com/v1

# Ollama with custom endpoint
python main.py --provider ollama --model llama3.1:8b --base-url http://localhost:11434

# With additional parameters
python main.py --provider anthropic --model claude-3-5-sonnet-20241022 --temperature 0.7 --max-tokens 2048

# Disable SSL verification (useful for self-signed certificates)
python main.py --preset llama3.1-8b --no-verify-ssl
python main.py --provider ollama --model llama3.1:8b --base-url https://my-ollama:11434 --no-verify-ssl

• --preset: Use a predefined model preset (mutually exclusive with --provider)
• --provider: Model provider (anthropic, openai, ollama) - requires --model

• --model: Model name (required when using --provider)
• --api-key: API key for the model provider
• --base-url: Base URL for the API (useful for custom OpenAI endpoints or Ollama)
• --temperature: Temperature for text generation (default: 0.0)
• --max-tokens: Maximum tokens in response (default: 1024)
• --no-verify-ssl: Disable SSL certificate verification (useful for self-signed certificates)

• ANTHROPIC_API_KEY: Your Anthropic API key
• OPENAI_API_KEY: Your OpenAI API key
• OPENAI_BASE_URL: Custom OpenAI-compatible endpoint (optional)
• OLLAMA_BASE_URL: Ollama server URL (default: http://localhost:11434)

1. Get an API key from Anthropic Console
2. Set the environment variable: export ANTHROPIC_API_KEY="your-key"
3. Available models: claude-3-5-sonnet-20241022, claude-3-haiku-20240307, etc.

1. Get an API key from OpenAI Platform
2. Set the environment variable: export OPENAI_API_KEY="your-key"
3. Available models: gpt-4-turbo-preview, gpt-3.5-turbo, etc.

1. Install Ollama from ollama.com
2. Start the Ollama server: ollama serve
3. Pull a model: ollama pull llama3.1:8b
4. Available models: Any model supported by your Ollama installation

The agent comes with these built-in tools:

• read_file: Read the contents of a file
• list_files: List files and directories in a path

To add custom tools, implement the ToolDefinition interface:

from agent import ToolDefinition

class MyCustomTool(ToolDefinition):
    def name(self) -> str:
        return "my_tool"

    def description(self) -> str:
        return "Description of what my tool does"

    def input_schema(self) -> dict:
        return {
            "type": "object",
            "properties": {
                "param": {"type": "string", "description": "Parameter description"}
            },
            "required": ["param"]
        }

    def execute(self, **kwargs) -> str:
        # Tool implementation
        return "Tool result"

Then add it to your agent:

from agent import MultiModelAgent, ModelFactory, ModelPresets, ModelProvider
from tools import ReadFileTool, ListFilesTool

# Create model
model = ModelFactory.create_model(ModelPresets.CLAUDE_3_5_SONNET)

# Add tools including your custom tool
tools = [ReadFileTool(), ListFilesTool(), MyCustomTool()]

# Create agent
agent = MultiModelAgent(model, ModelProvider.ANTHROPIC, get_user_message, tools)

You: List the files in the current directory
Agent: [uses list_files tool]

You: Read the contents of main.py
Agent: [uses read_file tool]

You: Analyze the structure of this Python project
Agent: [uses list_files to explore, then read_file to examine key files]

• ModelFactory: Creates model instances for different providers
• MultiModelAgent: Main agent class handling conversation and tool execution
• ToolDefinition: Interface for implementing custom tools
• Provider-specific handling: Different tool calling formats for each provider

1. Fork the repository
2. Create a feature branch
3. Add your changes with tests
4. Submit a pull request

MIT License - see LICENSE file for details.