penguin-ai 0.4.0

Penguin: A modular, extensible AI coding agent and software engineer with its own execution environment.

ooooooooo.                                                 o8o              
`888   `Y88.                                               `"'              
 888   .d88'  .ooooo.  ooo. .oo.    .oooooooo oooo  oooo  oooo  ooo. .oo.   
 888ooo88P'  d88' `88b `888P"Y88b  888' `88b  `888  `888  `888  `888P"Y88b  
 888         888ooo888  888   888  888   888   888   888   888   888   888  
 888         888    .o  888   888  `88bod8P'   888   888   888   888   888  
o888o        `Y8bod8P' o888o o888o `8oooooo.   `V88V"V8P' o888o o888o o888o 
                                   d"     YD                                
                                   "Y88888P'                                
                                         

To quickly understand the codebase, DeepWiki is recommended (note: it's ~90% accurate)

Table of Contents

• Overview
• Quick Start
• Development Status
• Key Features
• Installation
• Usage
• Architecture
• Contributing
• Support & Help
• License
• Acknowledgments

Overview

Penguin is a modular, open-source AI software engineer that combines autonomous code generation with project management, task coordination, and multi-agent orchestration to take projects from spec to implementation.

Quick Start

# Install with CLI tools (recommended for most users)
pip install penguin-ai

# Optional: Install with web interface 
pip install penguin-ai[web]

# Set up your API key (OpenRouter recommended)
export OPENROUTER_API_KEY="your_api_key"  # On Windows: set OPENROUTER_API_KEY=...

# Run the setup wizard
penguin config setup

# Start using Penguin
penguin              # Interactive CLI chat
penguin-web          # Web API server (if [web] installed)

Quick Python Usage

from penguin import PenguinAgent

# PenguinAgent handles PenguinCore setup, workspace defaults, and project docs
with PenguinAgent() as agent:
    response = agent.chat("Summarize the current task charter")
    print(response["assistant_response"])

    # Register a specialized sub-agent if desired
    agent.register_agent("qa", system_prompt="You are the QA reviewer.")

PenguinAgent wraps the full PenguinCore stack (conversation manager, tool manager, project docs autoloading) with a synchronous API so you can script multi-agent workflows without reimplementing orchestration.

Penguin

Penguin is a modular, extensible AI coding assistant powered by LLMs. It functions as an intelligent software engineer that can assist with coding tasks while maintaining its own code execution, memory tools, and workspace environment.

It is designed for full-lifecycle software development. It goes beyond code generation by managing tasks, coordinating sub-agents, tracking project progress, and executing long-running objectives with minimal human oversight. Its architecture includes persistent memory, a rich toolchain, CLI and Web interfaces, and an SQLite-backed project management system. Penguin enables scalable, intelligent workflows across complex codebases and development environments, making it a serious upgrade from prompt-based coding assistants.

Key Features

Core Orchestration

• Multi-agent runtime with planner/implementer/QA personas, lightweight "lite" agents, and per-agent routing through agent_id.
• Scoped sub-agent delegation that inherits context, tools, and memory while enforcing token and permission boundaries.
• Engine-managed reasoning loop that powers chat and Run Mode with configurable iterations and pluggable stop conditions (token budget, wall clock, external callbacks).
• MessageBus and telemetry streams that tag every event with agent/channel metadata for CLI, TUI, web, and dashboard consumers.

Conversation & Memory Systems

• ConversationManager that blends session persistence, auto-save, context loading, checkpoints, and snapshot/restore support.
• ContextWindowManager with category-based token budgets, multimodal trimming, and live usage reporting to keep histories within model limits. This allows for theoretically infinite sessions.
• Shared memory layer with declarative notes, summary notes, and retrieval backed by SQLite plus pluggable vector providers (FAISS, LanceDB, Chroma, others).

Development Tools & Workspace Automation

• Workspace-aware toolchain that respects project/workspace roots for file edits, diffs, pattern-based refactors, and repository operations.
• LLM-driven scaffolding for code, documentation, tests, and refactoring guidance.
• Analysis and execution utilities such as AST inspection, dependency mapping, linting, notebook-based execution, grep/workspace search, and Perplexity/web search integrations.
• Browser automation via headless navigator and PyDoll tools for scripted browsing and capture, plus repository helpers to scaffold PRs, manage branches, and push changes.

Project & Task Management

• SQLite-backed ProjectManager with ACID transactions, dependency graphs, resource budgets, execution tracking, and event bus integration.
• CLI, Python, and web surfaces for project/task CRUD, status tracking, budgeting, and autonomous execution via Run Mode.

Interfaces & Integrations

• Rich CLI with interactive TUI, setup wizard, and >20 commands covering projects, memory, and tooling.
• Async Python client (PenguinClient) that offers streaming chat, checkpoint workflows, model switching, and multi-agent routing.
• FastAPI web server with REST + WebSocket streaming endpoints and an embeddable PenguinAPI for custom applications (agent spawn/pause/delegate, conversation history, telemetry).
• Dashboard hooks and telemetry endpoints for observability integrations.

Model & Provider Support

• Native and gateway adapters for OpenAI, Anthropic, OpenRouter, and LiteLLM-supported backends (Azure, Bedrock, DeepSeek, Ollama, and more).
• Runtime model/provider switching with layered configuration, capability detection, and multimodal (vision/image) support.
• Provider-aware token counting, cost reporting, and budgets exposed through the engine and telemetry APIs.

Performance, Diagnostics & Governance

• Fast-startup path with lazy tool loading, deferred memory indexing, and background workers to reduce cold-start latency.
• Structured diagnostics covering startup profiling, operation timing, token usage, and error tracing.
• Telemetry collector aggregating message, agent, task, and token metrics for dashboards and alerting.
• Configurable logging, retries, and graceful error recovery across subsystems.

Data & Context Ingestion

• Context loader and cataloging pipeline for PDFs, docs, and workspace artifacts so agents can ground responses in project materials.
• Memory indexing with semantic search and declarative knowledge capture integrated into the conversation loop.

Extensibility & Configuration

• Tool registry and plugin architecture for declarative or dynamic tool registration without patching core modules.
• Comprehensive configuration surface through config.yml, environment variables, and CLI helpers (project/workspace roots, model defaults, streaming controls).
• Event bus and MessageBus hooks for integrating custom services with agent lifecycle events, tool invocations, and telemetry streams.

Prerequisites

• Python 3.9+ (3.10+ recommended for best performance)
• Valid API key(s) for your chosen AI model provider(s)
• UV package manager (optional, for development)

Installation

Recommended: PyPI Installation

# Core installation (includes CLI tools)
pip install penguin-ai

# With web interface
pip install penguin-ai[web]

# With memory providers
pip install penguin-ai[memory_faiss]    # FAISS + sentence-transformers  
pip install penguin-ai[memory_lance]    # LanceDB
pip install penguin-ai[memory_chroma]   # ChromaDB

# Full installation (all features)
pip install penguin-ai[all]

Development Installation

For contributing or using the latest features:

git clone https://github.com/maximooch/penguin.git
cd penguin/penguin
pip install -e .                        # Editable install
# OR with UV (faster)
pip install uv && python uv_setup.py    # Automated UV setup

Available Extras

Extra	Description
[web]	FastAPI server + WebSocket support
[memory_faiss]	FAISS vector search + embeddings
[memory_lance]	LanceDB vector database
[memory_chroma]	ChromaDB integration
[browser]	Browser automation (Python 3.11+ only)
[all]	Everything above

Usage

# Interactive chat
penguin

# Run setup wizard
penguin config setup

# Project management
penguin project create "My Project"
penguin project task create PROJECT_ID "Task description"

# Web API server (requires [web] extra)
penguin-web

For detailed usage, see the documentation.

Common In‑Chat Commands

When running penguin interactively, you can use slash‑style commands to control models, streaming, context, checkpoints, and run mode. Type /help in chat to see them all. A few useful ones:

/models                 # Interactive model selector
/model set <MODEL_ID>   # Set a specific model (e.g., openrouter/anthropic/...)
/stream on|off          # Toggle token streaming

# Checkpoints & branches
/checkpoint [name] [description]  # Save a conversation checkpoint
/checkpoints [limit]               # List checkpoints
/rollback <checkpoint_id>         # Restore to a checkpoint
/branch <checkpoint_id> [name]    # Branch a new convo from a checkpoint

# Context window & diagnostics
/truncations [limit]     # Recent context trimming events
/tokens|/tokens detail   # Token usage summary / details

# Context file helpers
/context add <glob>      # Copy files into workspace context
/context list|clear      # Inspect or clear context files
/context write|edit|remove|note   # Manage context artifacts

# Run Mode
/run task "Name" [desc]        # Run a specific task
/run continuous ["Name" [desc]] # Continuous Run Mode (alias: --247)

Architecture

Penguin follows a layered architecture that keeps reasoning, memory, tooling, and delivery surfaces loosely coupled while sharing telemetry and configuration.

Runtime Flow

1. Interfaces (CLI/TUI, Python client, web API) collect user prompts or task requests.
2. PenguinCore coordinates global configuration, instantiates shared services, and routes work into the Engine.
3. The Engine runs the reasoning loop: it prepares conversation state, chooses the right agent (planner, implementer, QA, lite agent, or sub-agent), requests completions through the APIClient, and dispatches tool invocations via ActionExecutor.
4. Results are persisted through the Conversation layer (ConversationManager, SessionManager, ContextWindowManager, CheckpointManager, SnapshotManager) and written back to interfaces via the MessageBus or streaming callbacks.
5. Project/task updates, memory notes, telemetry, and diagnostics are propagated to their dedicated subsystems for analytics and follow-up automation.

Key Subsystems

• Conversation Layer: ConversationManager manages sessions, context files, checkpoints, and snapshots. ContextWindowManager enforces category-based token budgets and trimming, while SessionManager persists conversations under the workspace. CheckpointManager/SnapshotManager provide branch/restore support.
• Engine & Multi-Agent Runtime: Engine orchestrates the reasoning loop, Run Mode, and stop conditions. It registers multiple agents (planner/implementer/QA/lite/sub-agents) via EngineAgent, integrates with MultiAgentCoordinator, and honors per-agent configuration.
• Tooling & Actions: ToolManager exposes 15+ built-in tools and lazy-loads heavy resources. ActionExecutor parses tool calls, NotebookExecutor runs code in an IPython kernel, and the plugin/registry system allows declarative or dynamic tool additions.
• Memory & Knowledge: Declarative notes, summary notes, and semantic search live in penguin.memory. Providers (SQLite, FAISS, LanceDB, Chroma, etc.) can be swapped or combined, and background indexers keep embeddings fresh.
• Project & Task Orchestration: ProjectManager provides ACID-backed CRUD, dependency graphs, budgeting, execution records, and event bus hooks. WorkflowOrchestrator, ProjectTaskExecutor, and ValidationManager coordinate complex project flows and Run Mode automation.
• Model & Provider Access: APIClient chooses the best adapter (native SDK, LiteLLM, OpenRouter) based on configuration. ModelConfig centralizes provider details, token budgets, reasoning modes, and streaming preferences.
• Interfaces: The Typer-based CLI powers both quick commands and the Textual TUI. The Python client (PenguinClient) wraps PenguinCore for async automation, and the FastAPI app exposes REST
  • WebSocket APIs plus an embeddable PenguinAPI class.
• Diagnostics & Telemetry: Profiling helpers, startup timing, message/agent telemetry, and the event bus feed dashboards, logs, and alerting. Logging is standardized via logging with subsystem-specific loggers.
• Configuration: config.yml, environment variables, and CLI flags converge into the Config object passed through PenguinCore. Execution roots, workspace overrides, model/provider choices, and feature flags are resolved hierarchically.

For deeper technical diagrams and API references, explore the documentation site.

Contributing

We welcome contributions! Penguin is open source and benefits from community involvement.

Quick Start for Contributors

# 1. Fork and clone the repository
git clone https://github.com/Maximooch/penguin.git
cd penguin/penguin

# 2. Set up development environment  
pip install -e .[dev,test]        # Install in development mode
pip install pre-commit && pre-commit install  # Set up code formatting

# 3. Run tests to ensure everything works
pytest tests/

# 4. Make your changes and test
# 5. Submit a pull request

How to Contribute

• 🐛 Bug Reports: Open an issue with details and reproduction steps
• 💡 Feature Requests: Discuss ideas before implementing
• 📖 Documentation: Help improve docs, examples, and guides
• 🧪 Testing: Add test coverage for new features and edge cases
• 🎨 UI/UX: Design improvements for CLI and web interfaces

Development Guidelines

• Follow PEP 8 style guidelines (enforced by black and ruff)
• Add docstrings for public functions and classes
• Include tests for new functionality
• Update documentation for user-facing changes
• Use semantic commit messages

For major changes, please open an issue first to discuss your approach.

Support & Help

Documentation & Resources

• Official Documentation - Complete user guide and API reference
• GitHub Discussions - Community Q&A and ideas
• Examples & Tutorials - Step-by-step guides

Issues & Bug Reports

• Report a Bug - Something not working?
• Request a Feature - Ideas for improvements

Project Status

• Current Version: v0.4.0
• Active Development: Phase 3 - Multi/Sub-agents, CLI/TUI refactor, GH integration, and achieving Claude Code parity then surpassing
• Stability: Core features stable, performance optimization in progress
• Python Support: 3.9+ (3.10+ recommended for best performance)

Changelog & Releases

• Release Notes - What's new in each version (right now it's just tags)
• (Soon) Development Blog - Technical deep-dives and progress updates

License

Open Source License

Penguin is free software: you can redistribute it and/or modify it under the terms of the GNU Affero General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Affero General Public License for more details.

Key points:

• You must disclose source code when you deploy modified versions
• Changes must be shared under the same license
• Network use counts as distribution
• Include copyright and license notices

For the complete license text, see:

• LICENSE file in this repository
• GNU AGPL v3 official text

Enterprise License

An enterprise license without the copyleft requirements is under consideration for organizations that need different licensing terms. This would allow:

• Proprietary modifications and integrations
• No obligation to share source code changes
• Commercial redistribution rights
• Priority support and consulting services

Interested in enterprise licensing? Please contact me at [e-mail] to discuss your requirements and explore available options.

Acknowledgments

Built upon insights from:

• CodeAct
• Claude-Engineer
• Aider
• RawDog