Algo - Voice-Enabled Task Graph Assistant

Algo is a voice-enabled personal AI assistant that helps you visualize and manage your daily routines as an interactive task graph. Using the OpenAI Realtime API for voice interactions, Algo understands your intent and automatically builds, validates, and maintains a directed graph of your tasks and routines.

Features

• 🎙️ Voice Interface - Natural conversations via OpenAI Realtime API
• 📊 Task Graph Visualization - Interactive directed graph showing your routines
• 🤖 Multi-Agent Architecture - Manager supervises Graph Agent with intelligent validation
• ✅ Automatic Corrections - Detects and fixes missing tasks, wrong names, and incorrect edge labels
• 🔄 Graph Operations - Rename tasks, insert between tasks, reorder routines
• 💾 Routine Persistence - Save and load your daily routines
• 🔐 Authentication - Secure login/register system

Installation

1. Clone the repository

   git clone https://github.com/jaseci-labs/Algo.git
   cd Algo

2. Set up your API key

   Create a .env file or set environment variable:

   export OPENAI_API_KEY='your-api-key-here'

3. Start the application

   jac start main.jac

4. Open your browser

   Navigate to http://localhost:8000

Authentication

• Register: Create a new account at /register
• Login: Access your account at /login
• Main App: Protected route at /app requires authentication

Architecture

Algo uses a Manager → Graph Agent multi-agent hierarchy:

System Architecture

Request Flow

Manager Agent

• Supervises and validates the Graph Agent's output
• Ensures user messages are correctly interpreted
• Handles corrections when Graph Agent fails
• Silent validator - never sends responses directly to users

Graph Agent

• Analyzes conversation intent using LLM semantic functions
• Extracts task names and builds graph relationships
• Updates the task graph with nodes (TaskState) and edges (TaskFlow)
• Performs corrections based on Manager validation feedback

Validation System

The Manager validates multiple operation types:

Operation	Description
Task Extraction	Ensures all tasks mentioned are captured
Task Rename	Validates task renames (old task removed, new task correct)
Insert	Validates task insertion position (before/after/between)
Reorder	Validates task reordering (no circular refs, correct order)
Edge Labels	Validates edge labels match user intent (conditional, parallel, sequential)

Supported Operations

• Rename: "rename MakeCoffee to GrabCoffee"
• Insert Before: "insert WashFace before CheckMessages"
• Insert After: "insert HaveBreakfast after MakeCoffee"
• Reorder: "have coffee before checking news"
• Conditional: "if raining, read book; otherwise, walk"
• Parallel: "have drink while chatting with buddies"

Edge Label Rules

Pattern	Label
Sequential (default)	"then", "afterwards"
Parallel tasks	"while", "at same time"
Conditional branches	"if X", "otherwise"
Conditional convergence	"either way"
Parallel convergence	"then", "afterwards"

Project Structure

Algo/
├── main.jac                    # Entry point
├── service/
│   ├── mainAppService.jac     # Service layer definition
│   └── mainAppService.impl.jac # Implementation (Graph Agent, Manager, helpers)
├── semantics/
│   └── semantics.jac           # LLM semantic functions (intent, extraction, validation)
├── api/
│   └── api.cl.jac              # Frontend API endpoints
├── hooks/
│   └── useMainApp.cl.jac       # React composition hooks
├── pages/
│   ├── MainApp.cl.jac          # Main app interface
│   ├── LoginPage.cl.jac        # Login page
│   └── RegisterPage.cl.jac     # Registration page
└── components/
    └── GraphViewer.cl.jac      # Graph visualization component

Development

# Install dependencies
pip install jaclang jac-client jac-scale byllm

# Run locally
jac start 

# Deploy to Kubernetes
jac start main.jac --scale

License

MIT