skewr

An AI Agent project built with Dart. Interact with LLMs through a CLI or Flutter app.

Demo

Project Structure

skewr/
├── chat_core/       # Core library - conversation, tool calling, event stream
├── chat_adapter/    # BLoC state management, adapter between Core and UI
├── skewr_cli/       # Command-line interface
└── skewr_app/       # Flutter app (future)

Dependency direction: skewr_cli / skewr_app → chat_adapter → chat_core

Getting Started

Prerequisites

• Dart SDK ^3.8.0
• A Portkey API key

Configuration

Create a local.properties file in the project root:

portkey-api-key=your-api-key
portkey-model=@provider-slug/model-name

Optional:

portkey-base-url=https://api.portkey.ai
portkey-max-tokens=1024

Configuration can also be set via environment variables (PORTKEY_API_KEY, PORTKEY_MODEL, PORTKEY_BASE_URL, PORTKEY_MAX_TOKENS) or CLI arguments (--api-key, --model, --base-url, --max-tokens).

Priority: CLI arguments > local.properties > environment variables.

Run the CLI

cd skewr_cli
dart pub get
dart run bin/skewr_cli.dart

Or with CLI arguments:

dart run bin/skewr_cli.dart --api-key your-key --model @openai/gpt-4o

Development

# chat_core
cd chat_core
dart pub get
dart run build_runner build
dart test

# chat_adapter
cd chat_adapter
dart pub get
dart test

# skewr_cli
cd skewr_cli
dart pub get
dart test

License

MIT