An AI goal-tracking journal that shows you how your habits reshape your body and mind — one honest check-in at a time.
📲 Klio is live on the App Store — iPhone, iPad and Mac, in English, Russian and Spanish.
Klio is a personal goal-tracking and diary application. You set a handful of everyday goals — quit smoking, run, read, sleep more — and check in each day. Instead of a bare streak counter, Klio uses Anthropic Claude to interview you when a goal is created, model the goal as a structured object (horizon, measure, direction, cadence), and then translate your consistency into a living picture of how your body and mind are actually changing.
Every day the app reacts to yesterday like an attentive coach, and a registry of "development spheres" (Lungs, Heart, Stamina, Sharpness, Mood…) rises and falls in response to real adherence — recovery is slow, relapses set you back, and inactivity quietly detrains you.
Everything the AI produces — questions, goal summaries, daily reactions, insights and sphere captions — is generated in the user's own language (Russian, English, Spanish), and the entire UI switches language at runtime.
- AI-guided goal creation — Claude parses a free-text goal into a taxonomy (eternal vs. situational, fact vs. quantitative, direction, baseline/target, cadence) and asks only the genuinely missing questions, one at a time.
- Flexible scheduling & streaks — daily / every-N-days / weekdays / times-per-week goals; streaks count only planned days, so a skipped off-day never breaks the chain.
- Daily check-ins & universal logs — mark goals done (binary) or record a value (quantitative), plus a single daily log of weight, sleep, energy and mood.
- Health-impact analytics — per-goal effect trajectories with milestone percentages, completion rates over 7/30/90 days, timelines for heatmaps, and weight/energy/mood trends.
- Development spheres — an AI-maintained 0–100 registry of body/mind/skill areas that each goal feeds into, moving realistically with adherence and relapse.
- AI diary & daily reactions — structured, multilingual coach cards (reaction / win / watch / tip / trend) plus weekly correlation insights ("on days you hit every goal, energy was 38% higher").
- Multilingual AI (ru / en / es) — one language directive drives every generated string; technical keys stay stable.
- Full UI localization with runtime switch — SwiftUI app re-renders instantly on language change.
- Social + email auth — Sign in with Apple and Google (server-side JWKS token verification) alongside classic email/password, all issuing app JWT access + refresh tokens.
flowchart LR
subgraph iOS["iOS App — SwiftUI (Swift 6)"]
UI["Today / Goals / Analytics / Diary / Profile"]
SA["Apple & Google Sign-In coordinators"]
LM["LocaleManager (ru / en / es)"]
API["APIClient + SessionStore (JWT)"]
end
subgraph Backend["FastAPI backend (Python 3.12, async)"]
R["/api/v1 routers\nauth · profile · goals · checkin · analytics · devices"]
SVC["Services\nAI · streak · schedule · insight · sphere · social-auth"]
SCH["APScheduler\n(daily reactions / sphere updates)"]
end
DB[("PostgreSQL 16\nSQLAlchemy 2.0 async · Alembic")]
Claude["Anthropic Claude API\n(claude-sonnet-4-6)"]
JWKS["Apple / Google JWKS"]
UI --> API
SA --> API
API -->|HTTPS + Bearer JWT| R
R --> SVC
SVC --> DB
SVC -->|prompts / JSON| Claude
SVC -->|verify id_token| JWKS
SCH --> SVC
The API is mobile-only: CORS is locked down, every mutating route is rate-limited via SlowAPI, and social identity tokens are verified against the providers' rotating JWKS before an app JWT is issued.
| Layer | Technology |
|---|---|
| Mobile | Swift 6, SwiftUI, iOS 17+ (strict concurrency), XcodeGen |
| Auth (client) | Sign in with Apple, Google OAuth (PKCE), Keychain session |
| Backend | Python 3.12, FastAPI, Uvicorn |
| Data | PostgreSQL 16, SQLAlchemy 2.0 (async / asyncpg), Alembic |
| Validation | Pydantic v2, pydantic-settings |
| Auth (server) | JWT (python-jose), passlib/bcrypt, JWKS verification for Apple/Google |
| AI | Anthropic Claude (claude-sonnet-4-6) via the anthropic async SDK |
| Scheduling | APScheduler |
| Hardening | SlowAPI rate limiting, locked-down CORS |
| Ops | Docker, Docker Compose, Caddy (reverse proxy / TLS) |
klio/
├── backend/ # FastAPI + PostgreSQL API
│ ├── app/
│ │ ├── main.py # App factory, CORS, rate limiter, /health
│ │ ├── config.py # Pydantic settings (env-driven)
│ │ ├── database.py # Async engine / session
│ │ ├── api/
│ │ │ ├── deps.py # Auth dependency (current user)
│ │ │ └── v1/ # auth · profile · goals · checkin · analytics · devices
│ │ ├── models/models.py # SQLAlchemy models (users, goals, spheres, logs…)
│ │ ├── schemas/ # Pydantic request/response schemas
│ │ └── services/ # ai · streak · goal_schedule · insight · sphere ·
│ │ # social_auth · plan · adapt · scheduler
│ ├── migrations/ # Alembic (001_initial … 009_profile_language)
│ ├── docker/Dockerfile
│ ├── docker-compose.yml # api + postgres (dev)
│ ├── pyproject.toml
│ └── .env.example
│
├── frontend/ # iOS app (Swift / SwiftUI)
│ ├── Klio.xcodeproj
│ ├── project.yml # XcodeGen spec
│ └── Sources/App/
│ ├── KlioApp.swift
│ ├── Core/ # Network · Storage · Localization · Router · Design
│ ├── Features/ # Auth · Onboarding · Home · Dashboard · GoalCreation ·
│ │ # Analytics · Profile · Splash
│ └── Resources/ # ru/en/es .lproj localization
│
├── docs/ # GOAL_DESIGN.md, screenshots/
├── SOCIAL_AUTH_SETUP.md # Apple & Google sign-in configuration guide
├── LICENSE
└── README.md
cd backend
# 1. Configure environment
cp .env.example .env
# then fill in: DATABASE_URL, SECRET_KEY, ANTHROPIC_API_KEY,
# APPLE_CLIENT_ID, GOOGLE_CLIENT_ID
# 2a. Run everything with Docker (api + postgres)
docker compose up --build
# API on http://localhost:8001 (Swagger at /docs)
# --- or run the API locally against your own Postgres ---
# 2b. Install and migrate
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
alembic upgrade head
uvicorn app.main:app --reloadcd frontend
open Klio.xcodeproj # then build & run the "Klio" scheme (iOS 17+)
# project.yml is the XcodeGen source of truth; run `xcodegen` to regenerate the project.Point AppConfig.apiBaseURL at your backend, and configure Apple/Google client IDs as
described in SOCIAL_AUTH_SETUP.md.
| Today | Create a goal | Analytics | Reflect (AI diary) |
|---|---|---|---|
 |
 |
 |
 |
Base path: /api/v1 (all non-auth routes require a Bearer access token).
| Method & Path | Purpose |
|---|---|
POST /auth/register · POST /auth/login · POST /auth/refresh |
Email/password auth, JWT access + refresh |
POST /auth/apple · POST /auth/google |
Verify provider id_token (JWKS) and issue app JWTs |
GET /profile · PUT /profile · DELETE /profile/me |
Read/update profile, delete account |
POST /goals/start → POST /goals/{id}/answer → POST /goals/{id}/confirm |
AI goal-creation dialog |
GET /goals · GET /goals/{id} · POST /goals/{id}/adapt · GET /goals/{id}/history · DELETE /goals/{id} |
Goal CRUD, adaptation, history |
GET /checkin/today · GET /checkin/{date} · POST /checkin |
Today's planned goals and daily check-in + log |
GET /analytics/goals/{id}/streak · .../effects · .../timeline |
Streaks, effect progress, heatmap timeline |
GET /analytics/daily-log/timeline · GET /analytics/spheres |
Weight/energy/mood trends, development spheres |
GET /analytics/insights · GET /analytics/insights/status · POST /analytics/insights/refresh |
AI insights (cached / on-demand) |
POST /devices/token |
Register APNs device token |
GET /health |
Liveness probe |
Interactive OpenAPI docs are served at /docs when the backend is running.
The core product is implemented end to end: async FastAPI backend with 9 Alembic migrations, email + Apple + Google auth, AI-driven goal creation, flexible scheduling and streaks, daily check-ins, health-impact analytics, development spheres, multilingual AI diary/insights, and a Swift 6 SwiftUI client with runtime localization.
Roadmap items still open (see backend/ROADMAP.md):
- Push notifications over APNs (device tokens are already captured) — check-in reminders and milestone/hint pushes.
- Broader automated test coverage (unit for streak/schedule/insight services, integration for every endpoint) and CI/CD via GitHub Actions.
MIT — see LICENSE. Copyright (c) 2026 Egor Fomenko.