mindful-agents

A powerful toolkit for interacting with large language models featuring vision capabilities and customizable agents.

Installation

pip install mindful-agents

Key Features

• 🤖 Multimodal Conversations
  • Text and image-based chat
  • Support for multiple images
  • Customizable system prompts
• 🔄 Flexible Integration
  • Interactive CLI chat
  • REST API server
  • Python library
• 💾 History Management
  • Save conversations (JSON/TXT/Markdown)
  • Load and continue chats
  • Organized by date and session

Usage

Python Library

from mindful_agents import MindfulAgents

# Initialize
mindful = MindfulAgents(
    mode='default', # Mode (default/chat/api)
    log_on=True, # Enable logging
    log_to='logs', # Log directory
    model='omni', # Model selection
    save_to='outputs', # History save path
    save_as='json', # Save format (json/txt/md)
    timeout=60 # Request timeout
)

# Text chat
response, history = mindful.get_completions(
    prompt="Your question here",
    agent='default',          # Agent type
    instruction=None,         # Custom system prompt
    history=None,            # Optional chat history
    chat_id=None             # Optional chat ID
)

# Image analysis
response, history = mindful.get_completions(
    prompt="Analyze this image",
    image_path="image.jpg", # Single image
    # OR
    image_path=["img1.jpg", "img2.jpg"],  # Multiple images
    agent='default', # Agent type
    instruction=None, # Custom system prompt
    history=None, # Optional chat history
    chat_id=None # Optional chat ID
)

# Load chat history
history = mindful.load_history("path/to/history.json")

Interactive CLI

Start the chat interface:

from mindful_agents import MindfulAgents
MindfulAgents(mode='chat')
# OR
mindful = MindfulAgents()
mindful.start_chat(
    agent='default', # Agent type
    instruction=None # Custom system prompt
)

Available commands:

• /exit - Exit chat
• /reset - Reset conversation
• /image "path" "question" - Send image
• /image ["path1", "path2"] "question" - Send multiple images
• /instruction "new prompt" - Change system prompt
• /load "history.json" - Load chat history
• /help - Show commands

REST API

Start the Flask API server:

from mindful_agents import MindfulAgents
MindfulAgents(mode='api')
# OR
mindful = MindfulAgents()
mindful.start_api(
    host="0.0.0.0", # Server host
    port=5000, # Server port
    debug=False # Enable debug mode
)

API Endpoints

• POST /v1/api/get/completions

Configuration

Save Formats

• json (default) - Complete conversation data
• txt - Plain text format
• md - Markdown format with images

Chat History

Chat histories are automatically saved and organized:

{save_to}/
  └── YYYY-MM-DD/
      └── YYYYMMDD_HHMMSS_UUID8.{format}

License

See LICENSE for details.