PaperCLI v2.1.0 - Reasoning Models & Service Architecture

🧠 OpenAI Reasoning Models Support

This release adds full support for OpenAI's reasoning models (o1, o3-mini, etc.) with configurable thinking content display and improved service architecture.

✨ Key Features

• Reasoning Model Integration – Native support for OpenAI reasoning models with visible thinking process
• ChatService Extraction – Refactored chat logic into dedicated ChatService for better separation of concerns
• Loading Animations – Visual feedback during chat response generation
• Configurable Reasoning Effort – Choose from minimal/low/medium/high effort levels for reasoning models

⚙️ Configuration

New environment variables for reasoning models:

• OPENAI_SHOW_THINKING – Display thinking content for reasoning models (true/false, default: false)
• OPENAI_REASONING_EFFORT – Set reasoning effort level (minimal/low/medium/high, default: medium)

Example:

export OPENAI_MODEL="o3-mini"
export OPENAI_SHOW_THINKING=true
export OPENAI_REASONING_EFFORT=medium

Configure via UI: Use /config command and navigate to the LLM tab for interactive settings.

🏗️ Architecture Improvements

• ChatService – Centralized chat operations including:
– OpenAI client initialization
– Token estimation with tiktoken
– PDF text extraction and cleaning
– Paper context building for LLM
– Streaming response handling with callbacks

• Callback Architecture – Replaced message-based streaming with direct callback functions for cleaner, more maintainable code

• Dialog Simplification – ChatDialog now focuses on UI rendering while ChatService handles business logic

📦 Installation

# Via pipx (recommended)
pipx install git+https://github.com/SXKDZ/papercli.git@v2.1.0

# Via pip
pip install git+https://github.com/SXKDZ/papercli.git@v2.1.0

# Update existing installation
pipx reinstall git+https://github.com/SXKDZ/papercli.git@v2.1.0

🔗 Full Changelog

v2.0.4...v2.1.0