Release Notes v0.2.2

Release Notes - v0.2.2

Release Date: December 2, 2024

What's New

Multimodal Conversation Support

Conversation now supports multimodal inputs through the attachments parameter. Send images, audio, video, and documents alongside your text messages.

Key Features:

• Flexible Input Formats: All attachment types accept local file paths, remote URLs, or base64-encoded data URIs
• Rich Media Support:
  • Images (JPEG, PNG, GIF, WEBP)
  • Audio (WAV, MP3, OGG)
  • Video (MP4, WEBM)
• Document Processing:
  • PDF documents
  • Word files (DOC, DOCX)
  • Excel spreadsheets (XLS, XLSX)
  • CSV data files
  • Plain text and Markdown files
• Web Content: Analyze web pages via URL links

Usage Example:

from chak import Conversation, Image, PDF

conv = Conversation("openai/gpt-4o", api_key="YOUR_KEY")

# Analyze an image
response = await conv.asend(
    "What's in this image?",
    attachments=[Image("photo.jpg")]  # Supports local path, URL, or base64
)

# Process a document
response = await conv.asend(
    "Summarize this report",
    attachments=[PDF("report.pdf")],
    timeout=120
)

# Multiple attachments
response = await conv.asend(
    "Compare these images",
    attachments=[
        Image("https://example.com/img1.jpg"),
        Image("./local/img2.png")
    ]
)

See Examples:

• Image processing: examples/multimodal_chat_image.py
• Document analysis: examples/multimodal_chat_documents.py

Documentation:

• Updated README with comprehensive multimodal support section
• Added usage examples for all supported file types

Technical Details

• Attachment Classes: New attachment types exported from chak package: Image, Audio, Video, PDF, DOC, Excel, CSV, TXT, Link
• MimeType Support: Explicit MIME type specification available via MimeType enum
• Custom Readers: Built-in readers for all document types, with support for custom reader functions
• Streaming Compatible: Multimodal inputs work seamlessly with streaming responses

Notes

• Not all LLM providers support all modalities. Check your provider's documentation for capabilities
• Large files may require longer timeouts. Use the timeout parameter when needed
• Both send() and asend() support attachments, but async is recommended for better performance with large files

Installation

Update to the latest version:

pip install --upgrade chakpy

# With all optional dependencies
pip install --upgrade chakpy[all]

---

For full details, see the updated README.