Release Notes for 0.5.0
Overview
This release introduces a set of substantial new features and architectural improvements. The focus is on practical automation, multimodal input, and extensibility for technical workflows.
Notable Features
Agent Workflow
A new agent workflow enables the program to execute shell commands iteratively in response to model output. The agent operates within a user-specified directory, processes command output, and can perform multi-step tasks. The number of actions is limited by a configurable parameter (default: 10). At the end of each session, a summary of executed commands is provided.
Example usage:
cgip agent ./project "analyze the codebase and suggest optimizations"
cgip agent ./app "create a basic Node.js Express app"
cgip agent ./src "find and fix compilation errors" --input main.rs
Image Analysis
The image subcommand supports analysis of visual content using OpenAI vision models. This includes document processing, code screenshot analysis, and user interface review. The tool selects an appropriate model for image tasks automatically.
Example usage:
cgip image --file invoice.png "Extract all line items and calculate totals"
cgip image --file screenshot.png "Describe the interface and suggest improvements"
Text-to-Speech
A tts subcommand is provided for text-to-speech conversion using OpenAI models. Multiple voices, quality levels, and audio formats are supported. Input can be provided via stdin or as arguments.
Example usage:
cat README.md | cgip tts --voice nova --output narration.mp3
cgip tts --voice echo --speed 0.8 "Automated build system announcement"
Embedding Generation
The embedding subcommand generates semantic embeddings for text, suitable for search and similarity applications.
Example usage:
find docs/ -name "*.md" -exec cgip embedding --output {}.vec {} \;
echo "machine learning algorithms" | cgip embedding --output query.vec
Architectural Changes
- The ChatGPT module has been refactored into separate components for API interaction, message handling, response parsing, and request construction.
- Improved compatibility with a broader range of OpenAI-compatible endpoints.
- Enhanced error reporting and updated dependencies, including base64 0.22.1.
Other Changes
- License changed from GPL-2.0 to BSD-3-Clause.
- Documentation expanded for all subcommands, with updated man pages and clarified API compatibility.
Summary
Version 0.5.0 extends the program’s capabilities from conversational interaction to practical automation and multimodal processing. The agent workflow, image analysis, text-to-speech, and embedding generation features are intended for integration into technical and command-line workflows. The release includes improvements in error handling, safety, and documentation.
To upgrade:
cargo install cgip
Refer to the documentation for further details.