Add Anthropic/OpenAI Provider Support

Added

• OpenAI Provider Support: Full integration with OpenAI API through dialectus-engine 0.4.0

  • Configure debates using OpenAI models (GPT-4, GPT-3.5, etc.)
  • Automatic API key validation at startup
  • Support for OPENAI_API_KEY environment variable

• Anthropic Provider Support: Full integration with Anthropic Claude models through dialectus-engine 0.4.0

  • Configure debates using Claude models (Claude 3.5 Sonnet, Claude 3 Opus, etc.)
  • Automatic API key validation at startup
  • Support for ANTHROPIC_API_KEY environment variable
  • Added example configuration for Anthropic debates

• Provider Key Validation: Enhanced startup checks to validate API keys before running debates

  • Clear error messages when required API keys are missing
  • Prevents debates from failing midway due to authentication issues

Changed

• Updated Engine Dependency: Now requires dialectus-engine >= 0.4.0

  • Benefits from improved provider architecture
  • Access to new OpenAI and Anthropic providers
  • Enhanced cost tracking and rate limiting

• Configuration Improvements: Better handling of provider-specific configuration

  • More robust config validation
  • Clearer error messages for misconfigured providers

• Documentation: Significantly expanded README with provider setup instructions

  • Detailed setup guides for all supported providers (Ollama, OpenRouter, OpenAI, Anthropic)

• Virtual Environment: Switched to .venv as the standard virtual environment directory

  • Aligns with modern Python tooling conventions (uv, etc.)

Fixed

• Character Encoding: Fixed cp1252 encoding issues on Windows systems
  • Improved handling of Rich library output on different platforms

Testing

• Enhanced Test Coverage: Added new tests for provider validation
  • Tests for API key validation logic
  • Tests for provider-specific configuration handling

v0.2.1

Overview

This release focuses on improving type safety, code quality, and maintainability through comprehensive refactoring. The CLI now uses Pydantic models throughout for better validation and type checking, with enhanced database interactions and improved presentation layer.

Breaking Changes

None - this is a backwards-compatible patch release.

New Features

Type-Safe Database Models

• New db_types.py module with comprehensive Pydantic models for all database interactions
• Type-safe row models: DebateRow, DebateMessageRow, JudgeDecisionRow, CriterionScoreRow, EnsembleSummaryRow
• Enhanced models: DebateWithMessages, JudgeDecisionWithScores, TranscriptData
• Custom exceptions: DatabaseError, DebateNotFoundError, JudgeDecisionNotFoundError, EnsembleSummaryNotFoundError

Improved Provider Detection

• Auto-detection of model providers in list-models command
• No longer requires manual provider specification for common models

Improvements

Type Safety

• Migrated TypedDict classes to Pydantic models throughout codebase
• Added strict type checking for all database operations
• Improved type annotations across all modules
• Better handling of optional fields with proper None checks

Database Layer (database.py)

• Refactored to return Pydantic models instead of dictionaries
• Type-safe constructors for all database operations
• Improved error handling with specific exception types
• Better separation of concerns with dedicated row types

Presentation Layer (presentation.py)

• Refactored to work with Pydantic models
• Improved dot-notation access for cleaner code
• Enhanced serialization/deserialization for display
• Better handling of ensemble summary data

Runner (runner.py)

• Type-safe model handling throughout debate execution
• Improved EnsembleResult integration from engine v0.3.1
• Better error handling and validation
• Cleaner async context management

Code Quality

• Extracted magic numbers to named constants
• Improved Click parameter naming for clarity
• Simplified context managers
• Better exception specificity
• Consistent code formatting with Ruff

Testing

Enhanced Test Coverage

• Updated all tests for Pydantic model assertions
• Better test isolation with improved fixtures
• More specific exception testing
• Refactored test utilities for maintainability

Dependencies

Updated

• dialectus-engine upgraded to >=0.3.1
  • Includes EnsembleResult improvements
  • Enhanced Pydantic type support
  • Better type safety across engine API

Developer Experience

Improved IDE Support

• Better autocomplete with Pydantic models
• Clearer type errors from Pyright
• Enhanced documentation through type hints
• More discoverable API surface

Code Organization

• Clearer separation between data models and business logic
• Better module organization with dedicated types file
• More consistent naming conventions
• Improved code documentation

Migration Notes

No migration required - existing databases continue to work unchanged. The new Pydantic models provide a type-safe layer over the existing SQLite schema.

Known Issues

None

---

For complete installation and usage instructions, see README.md.

For engine changes, see dialectus-engine v0.3.1 release notes.

Support the new type-safe event system

Updated for Type-Safe Engine API

This release updates the CLI to support the new type-safe event system introduced in dialectus-engine v0.3.0.

Changes

Engine Compatibility

• Updated to require dialectus-engine>=0.3.0
• Adopted new type-safe event callback system with enum-based event types
• Migrated from string-based event types to PhaseEventType and MessageEventType enums

Internal Improvements

• Type-Safe Callbacks: Event handlers now use strongly-typed enum values instead of string literals
• Enhanced Type Checking: Full support for Pyright/mypy type validation
• Better IDE Support: Improved autocomplete and inline documentation for event handling

Code Example

The CLI now uses the new type-safe API internally:

from dialectus.engine.debate_engine.types import (
    MessageEventType,
    PhaseEventType,
    PhaseStartedEventData,
    MessageCompleteEventData,
)

async def phase_callback(
    event: PhaseEventType,
    data: PhaseStartedEventData
):
    if event == PhaseEventType.PHASE_STARTED:
        # Handle phase start
        ...

async def message_callback(
    event: MessageEventType,
    data: MessageStartEventData | MessageCompleteEventData,
):
    if event == MessageEventType.MESSAGE_COMPLETE:
        # Handle message completion
        ...

User-Facing Features

No changes to CLI commands or user interface - all updates are internal improvements that maintain full backward compatibility with existing debate configurations and usage patterns.

Requirements

• Python >= 3.12
• dialectus-engine >= 0.3.0 (breaking change in engine)

Upgrade Notes

If you're upgrading from v0.1.x:

1. The CLI will automatically use the new engine when you update
2. No changes needed to your debate_config.json files
3. All CLI commands work exactly as before

# Upgrade to latest versions
pip install --upgrade dialectus-cli

# Or with uv
uv pip install --upgrade dialectus-cli

This release maintains CLI functionality while adopting the improved type safety of the underlying engine, providing a more robust foundation for future development.

Initial Release

Dialectus CLI v0.1.0 - Initial Release

First public release of the Dialectus command-line interface for running AI debates locally.

Overview

Dialectus CLI is a lightweight, local-first terminal application that brings AI-powered debates to your command line. Run debates between language models, judge them with AI critics, and save transcripts to a local database - all without needing a backend API.

Key Features

Core Functionality

• Local AI debates - Run debates completely offline using Ollama models
• Cloud model support - Use OpenRouter for access to GPT-4, Claude, and other cloud models
• Multiple debate formats - Oxford, Parliamentary, Socratic, and Public Forum styles
• AI judging - Evaluate debates with configurable judge models and criteria
• Rich terminal UI - Beautiful formatted output with tables, panels, and progress indicators
• SQLite storage - Save and review debate transcripts locally

Commands

• dialectus debate - Start a debate between AI models
• dialectus list-models - Show available models from Ollama and OpenRouter
• dialectus transcripts - View saved debate history

Configuration

• JSON-based configuration with full validation
• Supports both Ollama (local) and OpenRouter (cloud) providers
• Flexible judging criteria and model selection
• Environment variable support for API keys

Technical Highlights

• Python 3.13 with modern type annotations
• Direct engine integration - Imports dialectus-engine library directly
• Type-safe - Pydantic validation for all configuration
• Cross-platform - Works on Linux, macOS, and Windows
• Zero network dependencies - Runs completely offline with Ollama
• Modern async - Built with async/await throughout

Requirements

• Python 3.13 or higher
• Ollama (for local models) or OpenRouter API key (for cloud models)

Installation

pip install dialectus-cli

Quick Start

# Copy example config
cp debate_config.example.json debate_config.json

# Edit configuration
nano debate_config.json

# Run a debate
dialectus debate

Links

• GitHub: https://github.com/Dialectus-AI/dialectus-cli
• PyPI: https://pypi.org/project/dialectus-cli
• Engine: https://github.com/Dialectus-AI/dialectus-engine
• Issues: https://github.com/Dialectus-AI/dialectus-cli/issues

---

MIT License - Open source and free to use