kakumanu-harshitha/fitvision-ai

Purpose and Runtime Entry Points

AI Fitness Tracker is a full-stack fitness coaching system: browser-based pose/workout interaction, FastAPI APIs, persisted fitness/social data, and AI-assisted personalization/coaching. Backend starts at Ai_Fitness_Tracker/backend/app/main.py (uvicorn app.main:app); React starts from Ai_Fitness_Tracker/frontend/src/index.js and composes routes/UI in Ai_Fitness_Tracker/frontend/src/App.js.

Client Application and Live Workout Boundary

The React client owns interactive screens, camera presentation, local workout state, and API/WebSocket consumption. Live sessions center on Ai_Fitness_Tracker/frontend/src/screens/LiveWorkout.js and Ai_Fitness_Tracker/frontend/src/components/CameraView.js; shared state is provided through Ai_Fitness_Tracker/frontend/src/contexts/WorkoutContext.js. The browser-side MediaPipe workflow is the real-time pose boundary; it emits workout outcomes rather than requiring raw camera streams to be persisted by the backend.

HTTP API, Authentication, and Domain Routing

FastAPI versioned routers expose auth, workouts, plans, dashboards, social, AI, chatbot, hydration, and voice-facing capabilities under Ai_Fitness_Tracker/backend/app/api/v1/. Authentication/security dependencies and request policy are centralized in Ai_Fitness_Tracker/backend/app/api/dependencies.py, Ai_Fitness_Tracker/backend/app/core/security.py, and Ai_Fitness_Tracker/backend/app/core/middleware.py. JWT is the primary auth mechanism, with migration evidence for optional TOTP MFA.

Business Services and Persistence

Router handlers delegate core workout, activity, plan, and identity behaviors to service/repository layers: Ai_Fitness_Tracker/backend/app/services/workout_service.py, Ai_Fitness_Tracker/backend/app/services/activity_service.py, and Ai_Fitness_Tracker/backend/app/db/repositories/workout_repo.py. SQLAlchemy models and async database setup form the persistence boundary in Ai_Fitness_Tracker/backend/app/db/models.py and Ai_Fitness_Tracker/backend/app/db/database.py. Alembic migrations in Ai_Fitness_Tracker/backend/alembic/versions/ evolve the relational schema, including workout replay, plans, chatbot context, and social activity.

Computer-Vision Exercise Pipeline

The backend AI core supports capture-to-feedback processing: camera input, MediaPipe landmark detection, angle/feature derivation, temporal smoothing, exercise state logic, rep counting, and scoring. Key orchestration and boundaries are Ai_Fitness_Tracker/backend/app/core_ai/pipeline.py, Ai_Fitness_Tracker/backend/app/core_ai/pose/pose_detector.py, and Ai_Fitness_Tracker/backend/app/core_ai/processing/rep_counter.py. Exercise-specific rules reside in Ai_Fitness_Tracker/backend/app/core_ai/exercises/, keeping interpretable movement semantics separate from API and persistence code.

Personalization and AI Coaching

Personal targets and coaching combine stored user/workout context with lightweight ML and LLM-oriented coach logic. The Random Forest-related model path is represented by Ai_Fitness_Tracker/backend/app/core_ai/personalization_model.py and Ai_Fitness_Tracker/backend/app/core_ai/processing/ml_models.py; chatbot/AI endpoints are Ai_Fitness_Tracker/backend/app/api/v1/chatbot.py and Ai_Fitness_Tracker/backend/app/api/v1/ai.py. Coach implementations in Ai_Fitness_Tracker/backend/app/core_ai/coach/ produce guidance for plans, lifestyle, or workout feedback.

Real-Time, External Infrastructure, and Operations

WebSockets support real-time chat and social-duel/session updates via Ai_Fitness_Tracker/backend/app/api/v1/websockets.py and Ai_Fitness_Tracker/backend/app/websockets.py, consumed by Ai_Fitness_Tracker/frontend/src/hooks/useWebSocket.js. PostgreSQL is the intended production datastore; Redis is an additional runtime dependency managed through Ai_Fitness_Tracker/backend/app/core/redis.py. Material external/model services include MediaPipe, scikit-learn/joblib model artifacts, and an LLM provider configured through backend settings rather than embedded in the frontend.