pika

PIKA

Personal Intelligence Assistant - A voice-controlled AI assistant for your Mac desktop.

PIKA is a native macOS application that lets you interact with AI using your voice. Just say "Pika" to activate, ask questions, manage your calendar, save memories, and more.

Features

Core

• Voice Activation - Say "Pika" to wake the assistant, or use continuous listening mode
• Customizable Wake Words - Add your own wake words, train with voice samples, or type custom phrases
• AI-Powered Responses - Natural conversations powered by your choice of AI model (Gemini, GPT-4, Claude, Llama, and more via Requesty.ai)
• Natural Actions - Just speak naturally - PIKA understands intent and takes action

Productivity

• Google Calendar Integration - Add, edit, and delete calendar events with voice commands
• Reminders - Create reminders with multi-tier notifications (24h, 12h, 3h, 1h, 10min before, and at time)
• Memory System - PIKA remembers important information you tell it, with semantic search to recall relevant context

Information

• Weather Updates - Get current weather for any location
• Pokemon Search - Look up Pokemon stats and information

Entertainment

• Higher/Lower Game - Play a number guessing game with streak tracking

System Requirements

• macOS 10.15 (Catalina) or later
• Apple Silicon Mac (M1, M2, M3, or M4)
• Microphone access for voice commands
• Internet connection for AI responses

Note: PIKA currently only supports Apple Silicon Macs. Intel Mac support may be added in the future.

Installation

Download

1. Go to the Releases page

2. Download one of the following:

   • PIKA-*-lite.zip (~18MB) - Lighter download, requires Ollama installed separately
   • PIKA-*-full.zip (~150MB) - Includes bundled Ollama for offline AI embeddings

3. Extract the zip and drag PIKA.app to your Applications folder

Lite Version: Install Ollama

If you downloaded the lite version, install Ollama for the memory/embedding system:

# Install via Homebrew
brew install ollama

Or download from ollama.ai

PIKA will automatically start Ollama when needed.

Setup

When you first launch PIKA, a setup wizard will guide you through configuration:

1. AI Provider (Required)

PIKA uses Requesty.ai as a unified gateway to multiple AI providers.

1. Create a free account at requesty.ai
2. Generate an API key
3. Enter it in the setup wizard

Requesty supports models from:

• Google (Gemini 2.0 Flash, Gemini Pro)
• OpenAI (GPT-4o, GPT-4o Mini)
• Anthropic (Claude 3.5 Sonnet, Claude 3 Opus)
• Meta (Llama 3.1, Llama 3.2)
• And many more

2. Google Calendar (Optional)

To enable calendar integration:

1. Go to Google Cloud Console
2. Create a new project (or select existing)
3. Enable the Google Calendar API
4. Create OAuth 2.0 credentials (Desktop app type)
5. Enter the Client ID and Client Secret in PIKA's setup

3. Embedding Model

PIKA uses Ollama to run a local embedding model for the memory system. The setup wizard will:

• Start Ollama (bundled or system-installed)
• Pull the nomic-embed-text model (~274MB)

This runs locally - no data is sent to external servers for embeddings.

How It Works

Architecture

PIKA is built with:

• Wails - Go + WebView framework for native macOS apps
• Go backend - Handles AI communication, calendar sync, memory storage
• Embedded web UI - Modern interface with voice visualization
• SQLite database - Local storage for configuration, memories, and cache

All data is stored locally in ~/Library/Application Support/PIKA/.

Voice Commands

1. Wake Word: Say "Pika" to activate listening
2. Continuous Mode: Toggle continuous listening in settings
3. Speech Recognition: Uses your Mac's built-in speech recognition (Web Speech API)
4. Text-to-Speech: Responses are spoken aloud (can be disabled)

Actions

PIKA can understand your intent and take actions:

Memory System

PIKA uses vector embeddings to store and retrieve memories semantically:

• Important information is automatically extracted and saved
• When you ask questions, relevant memories are retrieved as context
• Embeddings are generated locally using Ollama (no data leaves your machine)

Development

Prerequisites

• Go 1.24 or later
• Wails CLI

# Install Wails
go install github.com/wailsapp/wails/v2/cmd/wails@latest

# Verify installation
wails doctor

Running Locally

# Clone the repository
git clone https://github.com/BasWilson/pika.git
cd pika

# Run in development mode (HTTP server)
make dev

# Or with hot reload
make install    # Install air for hot reload
make dev-server

Open http://localhost:8080 in your browser.

Building

# Build for Apple Silicon
make build

# Optionally bundle Ollama into the app
make bundle-ollama

# The app will be at: build/bin/PIKA.app

Project Structure

pika/
├── main.go              # Wails app entry point
├── internal/
│   ├── server/          # HTTP server & API routes
│   ├── ai/              # AI service (Requesty + Ollama)
│   ├── calendar/        # Google Calendar integration
│   ├── reminder/        # Reminders with scheduled notifications
│   ├── memory/          # Vector memory store
│   ├── actions/         # Action handlers (calendar, weather, games, etc.)
│   └── config/          # Configuration management
├── web/
│   ├── templates/       # HTML templates
│   └── static/          # CSS & JavaScript
└── build/               # macOS app bundle output

Privacy

• Local-first: All data stored in ~/Library/Application Support/PIKA/
• Embeddings: Generated locally via Ollama - no data sent externally
• AI Responses: Sent to your configured AI provider (via Requesty.ai)
• Calendar: Synced with Google Calendar if you enable integration
• No telemetry: PIKA does not collect usage data

Troubleshooting

Microphone not working

1. Open System Preferences > Privacy & Security > Microphone
2. Ensure PIKA has permission

Ollama not starting

If using the lite version:

# Check if Ollama is installed
which ollama

# Start manually if needed
ollama serve

Calendar not syncing

1. Check Google OAuth credentials in setup
2. Re-authorize by clicking "Connect Google Calendar" in settings

License

MIT License - See LICENSE for details.

Built with Wails and Requesty.ai