GitHubHelp

AsharibAli/MyTaskAI

Organize your life, your way | AI-Powered Task Management

★ 0Forks 0TypeScriptGitHub ↗Compare

README

MyTaskAI Logo

MyTaskAI — Organize your life, your way

AI-native to-do application with a conversational interface and a graphical UI. The chat agent uses LLM tool calling and an MCP-style tool layer to create, update, search, tag, and complete tasks on the user's behalf.

Architecture

┌──────────────────────────┐        HTTPS         ┌──────────────────────────┐
│   Next.js Frontend       │  ───────────────────▶│   FastAPI Backend        │
│   (Cloudflare Workers)   │                       │   (FastAPI Cloud)        │
│                          │                       │                          │
│   • Chat UI              │                       │   • Auth (JWT)           │
│   • Tasks UI             │                       │   • Tasks CRUD           │
│   • OAuth (optional)     │                       │   • Conversations        │
└──────────────────────────┘                       │   • Chat agent + tools   │
                                                   │   • Inline recurrence    │
                                                   └────────────┬─────────────┘
                                                                │
                                                                ▼
                                                        ┌──────────────┐
                                                        │  PostgreSQL  │
                                                        │  (managed)   │
                                                        └──────────────┘

Two deployable units, one managed database. The agent reaches LLMs through OpenRouter, so the model and provider are a one-line config change.

Repository layout

.
├── backend/        FastAPI service — deploys to FastAPI Cloud
│   ├── src/
│   │   ├── api/             REST endpoints (auth, tasks, chat, conversations, health, password reset)
│   │   ├── core/            config, database, security, logging, llm (OpenRouter client)
│   │   ├── mcp/             chat agent + tool definitions
│   │   ├── middleware/      request logging + error handling
│   │   ├── models/          SQLModel entities
│   │   ├── services/        business logic (tasks, auth, chat, conversations, email)
│   │   └── utils/           date parsing helpers
│   ├── alembic/             database migrations
│   ├── main.py              uvicorn entry point (re-exports src.main:app)
│   ├── pyproject.toml
│   └── .env.example
│
├── frontend/       Next.js app — deploys to Cloudflare Workers (via OpenNext)
│   ├── src/
│   │   ├── app/             routes (auth, chat, home, landing)
│   │   ├── components/      UI (chat, tasks, navigation, ui primitives)
│   │   ├── lib/             API client + utilities
│   │   └── types/           shared TS types
│   ├── public/_headers      Workers Static Assets caching headers
│   ├── package.json
│   ├── next.config.ts       wires OpenNext for `next dev`
│   ├── open-next.config.ts  OpenNext + Cloudflare config
│   ├── wrangler.jsonc       Workers deployment config
│   └── .env.example
│
├── specs/          spec-driven-development artifacts (spec, plan, tasks)
├── history/        prompt history records
└── .specify/       SDD templates and project constitution

Core features

  • Conversational AI interface — chat with an LLM agent that operates on your task list via tool calls. The model is reached through OpenRouter, so you can switch from openai/gpt-4o-mini to anthropic/claude-sonnet-4, google/gemini-2.5-flash, or any other supported model by changing one env var.
  • Graphical UI — full task list, filters, search, tagging, due-dates, priorities.
  • MCP-style tool layer — add_task, list_tasks, complete_task, update_task, delete_task, set_priority, add_tag, set_due_date, search_tasks, set_reminder, set_recurrence, etc.
  • Authentication — JWT auth, optional Sign-in-with-Google, password reset via Resend email.
  • Recurring tasks — when a recurring task is completed the next occurrence is created inline by the backend.

Local development

Prerequisites

  • Python 3.11+
  • Node.js 18+
  • A PostgreSQL database (local Postgres, Neon, Supabase, etc.)
  • An OpenRouter API key (sk-or-v1-…)
  • uv for the backend, pnpm for the frontend

Backend

cd backend
cp .env.example .env.local       # fill in DATABASE_URL, JWT_SECRET, OPENROUTER_API_KEY, ...
uv venv
uv pip install -e .
uv run alembic upgrade head      # run migrations
uv run uvicorn main:app --reload --host 0.0.0.0 --port 8000

API: http://localhost:8000 — interactive docs: http://localhost:8000/docs.

Frontend

cd frontend
cp .env.example .env.local       # set NEXT_PUBLIC_API_URL=http://localhost:8000
pnpm install
pnpm dev

App: http://localhost:3000.

Deployment

Backend → FastAPI Cloud

See https://fastapicloud.com/docs/getting-started/existing-project/

cd backend
fastapi deploy

Set the following environment variables in the FastAPI Cloud dashboard:

Variable Required Notes
DATABASE_URL yes Postgres connection string
JWT_SECRET yes 32+ random bytes
OPENROUTER_API_KEY yes One key for every supported provider via OpenRouter
CORS_ORIGINS yes Include the Cloudflare Workers domain (e.g. https://taskai-frontend.<account>.workers.dev or your custom domain)
LLM_MODEL no OpenRouter model slug — e.g. openai/gpt-4o-mini, anthropic/claude-sonnet-4, google/gemini-2.5-flash. Defaults to openai/gpt-4o-mini
OPENROUTER_BASE_URL no Defaults to https://openrouter.ai/api/v1
OPENROUTER_SITE_URL, OPENROUTER_APP_NAME no Optional ranking headers shown on the OpenRouter leaderboard
RESEND_API_KEY, RESEND_FROM_EMAIL, FRONTEND_URL no Required only for password-reset emails
LOG_LEVEL no Defaults to INFO

Frontend → Cloudflare Workers (via OpenNext)

The Next.js app is bundled for the Workers runtime by @opennextjs/cloudflare. Deploy from your machine with:

cd frontend
pnpm install
pnpm deploy             # runs `opennextjs-cloudflare build && opennextjs-cloudflare deploy`

Or wire Cloudflare Workers Builds (Git → CI). In the Cloudflare dashboard under Workers & Pages → taskai-frontend → Settings → Builds, set:

Field Value
Root directory frontend
Build command npx opennextjs-cloudflare build
Deploy command npx wrangler deploy
Install command blank (auto-detected from pnpm-lock.yaml)

Set the following as build variables under Settings → Variables and Secrets (NEXT_PUBLIC_* are inlined into the client bundle at build time):

Variable Required
NEXT_PUBLIC_API_URL yes — the FastAPI Cloud URL
NEXT_PUBLIC_GOOGLE_CLIENT_ID optional, for Google OAuth

After both are deployed, update CORS_ORIGINS on the backend to include the worker URL (https://taskai-frontend.<account>.workers.dev or your custom domain).

Switching LLM providers

The agent talks to models exclusively through OpenRouter. Provider choice is a single environment variable:

LLM_MODEL=openai/gpt-4o-mini             # default
LLM_MODEL=anthropic/claude-sonnet-4      # switch to Claude
LLM_MODEL=google/gemini-2.5-flash        # switch to Gemini
LLM_MODEL=meta-llama/llama-3.3-70b-instruct
LLM_MODEL=mistralai/mistral-large-latest

No code changes required. The full catalogue is at openrouter.ai/models. Centralized client construction lives in backend/src/core/llm.py; every other call-site just imports get_llm_client and LLM_MODEL from there.

API surface

Method Path Purpose
POST /api/auth/register Register a new account
POST /api/auth/login Login, returns JWT
GET /api/auth/me Current user profile
POST /api/auth/forgot-password Send reset code
POST /api/auth/reset-password Reset password with code
POST /api/auth/oauth OAuth login/sync (Google)
GET /api/tasks List tasks (filters, sort)
POST /api/tasks Create task
PUT /api/tasks/{id} Update task
PATCH /api/tasks/{id}/complete Toggle completion
DELETE /api/tasks/{id} Delete task
GET /api/tasks/search?q=... Search by keyword
POST /api/tasks/{id}/tags Add tag
DELETE /api/tasks/{id}/tags/{name} Remove tag
POST /api/tasks/{id}/reminder Set reminder time
POST /api/chat/chat Send message to agent
GET /api/conversations List conversations
GET /api/conversations/{id} Get conversation
GET /health/live Liveness probe
GET /health/ready Readiness probe (checks DB)

Full schema: http://localhost:8000/docs.

Spec-driven development

This project follows a spec-driven workflow. See specs/001-taskai-core-platform/ for the spec, plan, and task list. Project rules and templates live in .specify/ and CLAUDE.md.

License

MIT

Contributors

AsharibAli

Issues