GymLogAI

GymLogAI is a minimal multi-project skeleton for importing workout messages, parsing them into structured training history, and generating workout recommendations.

Architecture

• GymLogAI.Core: domain entities, enums, and primitive abstractions like IClock and IIdGenerator
• GymLogAI.Application: repository contracts, AI contracts, DTOs, and use-case handlers
• GymLogAI.Persistence: EF Core AppDbContext, PostgreSQL mappings, repositories, infrastructure services, and the first migration
• GymLogAI.AI: OpenRouter-ready options and stubbed AI implementations for parsing, recommendations, and embeddings
• GymLogAI.API: minimal API composition root with Swagger, import/recommend/history endpoints, and database migration on startup
• GymLogAI.TelegramBot: Telegram long-polling host with /start, /today, and text message import
• GymLogAI.Worker: background worker that polls pending telegram messages and parses them into workouts

Runtime Flow

1. API or Telegram bot imports a raw Telegram message into telegram_messages with Pending status.
2. Worker polls pending messages and runs ParseTelegramMessageHandler.
3. Parsed workouts are stored in workouts, workout_exercises, and exercise_sets.
4. API and Telegram bot can request workout recommendations from the application layer.

Key Endpoints

• POST /api/messages/import
• POST /api/workouts/recommend
• GET /api/workouts/history?userId=<guid>
• Swagger UI: /swagger

Database

• Provider: PostgreSQL
• Migration project: GymLogAI.Persistence
• Initial migration is included in GymLogAI.Persistence/Migrations
• Exercise includes an Embedding field stored as nullable real[]

Run

dotnet build
dotnet run --project GymLogAI.API/
dotnet run --project GymLogAI.TelegramBot/
dotnet run --project GymLogAI.Worker/
docker compose -f compose.yaml up --build

Configuration

• Connection string: ConnectionStrings:Postgres
• Telegram bot token: TelegramBot:BotToken
• OpenRouter settings: OpenRouter:*
• Worker polling settings: Worker:BatchSize, Worker:PollingIntervalSeconds