polimind

An extensible Python toolkit for AI capabilities. Its current public OpenAI interface lives under polimind.api.openai and preserves the existing chat, history, multimodal, and metrics workflow.

Features

• Centralized Client: Use a single Client to manage your connection and spawn multiple chat sessions.
• Variadic Chat Interface: Send multiple message components directly: chat.chat("Text", Image("local.png"), "More text").
• Node-Based History: Every interaction is a node with an incremental ID.
• Context Control: Activate or deactivate conversation nodes by their ID to manage the model's window.
• Enhanced Metrics: Track token usage, models, and active nodes per request with built-in Pandas integration.
• Multimodal Support: Handle local images (auto-base64 encoding) and remote URLs seamlessly.
• State Persistence: Save Chat states and load them via the Client to resume sessions.
• Deep Copy: Create independent clones of your chat session.

Installation

# Via Pip (once published)
pip install polimind

# From source
git clone https://github.com/artitzco/polimind.git
cd polimind
pip install .

Quick Start

from polimind.api.openai import Client, Image

# Initialize the main client
client = Client(api_key="your-api-key") # or use OPENAI_API_KEY env var

# Start a new chat session
chat = client.chat(model="gpt-4o-mini", system_prompt="You are a helpful assistant.")

# Basic Chat
response = chat.chat("Hello! What can you do?")
print(response)

# Multimodal Chat (Local and URL)
chat.chat(
    "Analyze these images:",
    Image("path/to/local_image.png"),
    "and also this one:",
    Image("https://example.com/image.jpg")
)

# Manage History
df_history = chat.history.to_dataframe()
chat.history.toggle(node_id=1, active=False) # Context management

# Persistence
chat.save("my_session.json")
new_chat_recovered = client.load_chat("my_session.json")

Project Structure

• src/polimind/: Root package for the project.
• src/polimind/api/openai/: OpenAI integration exposed by the package.
  • __init__.py: Contains the Client factory and exports.
  • core.py: Contains the Chat session manager.
  • history.py: Logic for node-based conversation management.
  • metrics.py: Token usage and request tracking.
  • content.py: Multimodal data builders (Images, etc.).
• notebooks/: Interactive guides and testing playgrounds.
• img/: Storage for sample assets.
• output/: Default location for saved session JSONs.

License

MIT