joao 0.1.0

A lightweight Python library for building chat agents with function calling capabilities

joao

A dual-purpose tool providing both a Python library for building chat agents and a ready-to-use CLI chat interface. Built with function calling capabilities and compatible with OpenAI-style APIs.

Features

• As a Library:

  • Simple and clean API for chat interactions
  • Support for function calling (tools)
  • Both synchronous and asynchronous APIs
  • Streaming support for real-time responses
  • Compatible with any OpenAI-style API (tested with Gemini 2.0 Flash)

• As a CLI:

  • Interactive chat interface
  • Single prompt mode
  • Rich markdown formatting for output
  • Temperature control for response randomness
  • Configurable system prompts
  • Multiple API provider support

• Minimal dependencies

Requirements

1. Python 3.7 or higher
2. An API key from one of these providers:
   • Gemini (Recommended): Get your API key from Google AI Studio
   • OpenAI: Get your API key from OpenAI Platform

Installation

1. Install the package:

   pip install joao
   

2. Set up environment variables:

   For Gemini (default):

   export OPENAI_API_KEY="your_gemini_api_key"
   

   For OpenAI:

   export OPENAI_API_KEY="your_openai_api_key"
   export OPENAI_BASE_URL="https://api.openai.com/v1"
   export OPENAI_MODEL="gpt-3.5-turbo"  # or your preferred model
   

Usage

Command Line Options

# Start interactive chat mode (with system prompt)
joao -s "You are a helpful assistant"

# Start chat mode without system prompt
joao

# Single prompt mode with system prompt
joao -s "You are a helpful assistant" "What is the capital of France?"

# Single prompt mode without system prompt
joao "What is the capital of France?"

# Enable streaming output
joao -s "You are a storyteller" --stream "Tell me a story"

# Set temperature (0.0-2.0, default: 0)
joao -s "You are a creative writer" -t 0.7 "Tell me a creative story"

# Show raw output without markdown formatting
joao -s "You are a helpful assistant" --raw "Hello"

# Use different API provider
joao -s "You are a helpful assistant" -e OPENAI "Using OpenAI API"

# Show debug info
joao -s "You are a helpful assistant" --debug "Test message"

Interactive Chat Commands

When in chat mode, you have access to these commands:

• /reset - Clear conversation history
• /reset <prompt> - Clear history and set new system prompt
• Ctrl+C - Exit chat session

Simple Chat API

from joao import Agent

# Create an agent with a personality
agent = Agent("You are Snoopy, the beloved Peanuts character")

# Get a response
response = agent.request("Who are your friends?")
print(response)

# Get a streamed response
for token in agent.request("Tell me a story", stream=True):
    print(token, end="", flush=True)
print()  # Final newline

Using Function Calling (Tools)

from joao import Agent

def drive_to(location: str):
    """Drive to the specified location"""
    print(f"Driving to {location}!")

# Create an agent with tools
agent = Agent("You are a helpful driver")
response = agent.request(
    "Can you drive me to San Francisco?", 
    tools=[drive_to],
    auto_use_tools=True  # Auto-execute any tool calls
)

Async Support

from joao import AsyncAgent
import asyncio

async def main():
    agent = AsyncAgent("You are a helpful assistant")
    response = await agent.request("Hello!")
    print(response)

    # With streaming
    async for token in agent.request("Tell me a story", stream=True):
        print(token, end="", flush=True)
    print()

asyncio.run(main())

Environment Variables

• OPENAI_API_KEY - Your API key
• OPENAI_BASE_URL - API endpoint (default: Gemini endpoint)
• OPENAI_MODEL - Model to use (default: gemini-2.0-flash)

You can prefix these with any string by using the -e flag, e.g., -e AZURE will look for AZURE_OPENAI_API_KEY