diff --git a/requirements_backlogr.md b/requirements_backlogr.md
new file mode 100644
index 0000000..ed99667
--- /dev/null
+++ b/requirements_backlogr.md
@@ -0,0 +1,347 @@
+# Backlogr — Requirements
+
+> **Concept:** “Letterboxd, but for video games” — a social platform to **log games you play**, **write reviews**, and **discover what to play next** with **AI-powered recommendations** and **vector (semantic) search**.
+
+---
+
+## Elevator pitch 
+
+Backlogr is a social app for gamers to track what they play, rate and review games, and follow friends’ activity.  
+It exists to make gaming journaling simple **and** help you find your next great game with AI—based on your tastes, not just what’s trending.
+
+---
+
+## Target audience
+
+- Gamers who want to **log** and **review** what they play (casual → hardcore)
+- People who like seeing **what friends are playing** and discussing it
+- Players who want **smarter recommendations** that match their preferences
+
+---
+
+## MVP feature scope 
+
+### Social + content
+- Follow system (follow/unfollow)
+- Activity feed (logs + reviews from followed users)
+- Likes on reviews
+- Comments on reviews
+
+### Logging + reviews
+- Personal library with statuses: **Playing / Played / Backlog / Wishlist / Dropped**
+- Rating scale in **0.5 increments**
+- Reviews (create/edit/delete) with optional **spoiler flag**
+- Game pages showing details + aggregated community activity
+
+### AI (core functionality)
+- AI “What should I play next?” recommendations (**core**)
+- AI review writing assistant (**secondary core**) to help draft/expand/clean up review text
+
+---
+
+## Game catalog source (IGDB)
+
+The app will use the **IGDB API** as the primary source of game data.
+
+Planned approach:
+- Add an internal “game import” flow:
+  - Search IGDB → select a game → store it in local DB (cache) with its IGDB id
+- Store common fields locally so the app stays fast and stable:
+  - title, summary/description, cover image URL, release date, platforms, genres, companies (optional)
+- When viewing games, prefer local cached data; refresh from IGDB as needed.
+
+---
+
+## Core use cases (user stories)
+
+### Account + roles
+- As a user, I can register/login/logout and manage my profile (display name, avatar, bio).
+- As an admin, I can moderate content (remove reviews/comments, and manage imported games if needed).
+
+### Discover + browse
+- As a user, I can browse/search games (initially using IGDB search; locally cached when imported).
+- As a user, I can open a game page to view details, community reviews, and my own history with the game.
+
+### Logging (my library)
+- As a user, I can add a game to my library with a status (Playing/Played/Backlog/Wishlist/Dropped).
+- As a user, I can set a rating (0.5 increments) and optional metadata:
+  - platform played on, hours played (optional), started/finished dates (optional), personal notes
+- As a user, I can filter/sort my library (status, rating, date updated, platform).
+
+### Reviews + ratings
+- As a user, I can write a review and mark spoilers.
+- As a user, I can edit/delete my review.
+- As a user, I can like reviews and comment on them.
+
+### Feed + follow
+- As a user, I can follow/unfollow other users.
+- As a user, I can see a feed of recent activity from people I follow (logs + reviews).
+
+---
+
+## AI + vector search
+
+### AI feature 1 (core) — “What should I play next?”
+- User clicks **Recommend** and the app returns a ranked list of suggested games.
+- Inputs to the recommendation:
+  - user’s logged games + ratings
+  - review text (optional, if available)
+- Output includes:
+  - recommended games
+  - short “why you might like it” explanation per result
+
+### AI feature 2 — Review writing assistant
+- In the review editor, user can:
+  - generate a first draft from bullet points
+  - rewrite for clarity/tone
+  - shorten/expand
+  - add spoiler-safe summary
+- The assistant must **never auto-post**; it only produces draft text the user can edit.
+
+### Vector search (semantic search)
+- User can search with natural language:
+  - “cozy story game with crafting and exploration”
+  - “hard roguelike with fast runs”
+- Results are driven by embeddings + vector similarity, not just keyword matching.
+
+### Implementation choice: Azure AI Search (vector search)
+- Store canonical data in Azure SQL.
+- Create an Azure AI Search index with:
+  - game text fields (title/summary/genres/platforms)
+  - an embedding vector per game (and optionally per review)
+- Use Azure AI (embeddings) to generate vectors and Azure AI Search to query by similarity.
+
+---
+
+## Tech stack
+
+### Front end
+- Vue 3
+- Vuetify 3
+- TypeScript (style guide enforced)
+- Axios (API calls)
+- Vitest (unit tests)
+- Responsive layout (Vuetify grid + breakpoints)
+
+### Back end
+- .NET 8 / ASP.NET Core Web API
+- C#
+- Entity Framework Core 8
+- ASP.NET Core Identity (JWT auth)
+- Role-based authorization (**User**, **Admin**)
+
+### DevOps / hosting
+- Azure Static Web Apps (frontend)
+- Azure App Service (API)
+- Azure SQL Database (primary DB)
+- Azure AI Search (vector search)
+- Azure AI (embeddings + chat/completions for assistant)
+- GitHub Actions CI/CD (build, test, deploy)
+
+---
+
+## High-level architecture
+
+1. **Vue SPA** calls API via Axios  
+2. **ASP.NET API** handles auth, validation, business logic (services), DTO mapping  
+3. **Azure SQL** stores users, games, logs, reviews, follows, comments, likes  
+4. **IGDB integration service** fetches and caches game metadata into SQL  
+5. **Embedding pipeline** generates vectors for games (and optionally reviews)  
+6. **Azure AI Search** performs semantic search + similarity retrieval  
+7. **AI recommend/assistant endpoints** use retrieved context to produce outputs
+
+---
+
+## Data model 
+
+### Naming rules
+- Primary key column naming: **TableNameId**
+- Foreign keys also use the target table name: **UserId**, **GameId**, **ReviewId**, etc.
+- For tables that act like a “join” (Follow/Like), we still use a **TableNameId** primary key and enforce uniqueness with constraints.
+
+### User (Identity + profile)
+> Identity provides the auth pieces (password hashes, security stamps, etc.). Backlogr adds profile fields on top.
+
+- **User**
+  - **UserId** (PK) *(Identity key; stored as string or Guid depending on implementation)*
+  - Email
+  - Username
+  - DisplayName
+  - FirstName *(optional)*
+  - LastName *(optional)*
+  - AvatarUrl *(optional)*
+  - Bio *(optional)*
+  - CreatedAt
+
+### Core content
+- **Game**
+  - **GameId** (PK)
+  - IgdbId (unique)
+  - Title
+  - Summary
+  - CoverUrl
+  - ReleaseDate *(optional)*
+  - Genres *(string or separate table, TBD)*
+  - Platforms *(string or separate table, TBD)*
+  - LastSyncedAt
+
+- **GameLog**
+  - **GameLogId** (PK)
+  - UserId (FK → User)
+  - GameId (FK → Game)
+  - Status *(Playing/Played/Backlog/Wishlist/Dropped)*
+  - Rating *(0.0–5.0 in 0.5 steps, optional)*
+  - Platform *(optional)*
+  - Hours *(optional)*
+  - StartedAt *(optional)*
+  - FinishedAt *(optional)*
+  - Notes *(optional)*
+  - UpdatedAt
+  - **Constraint:** unique (UserId, GameId) so a user has one log per game
+
+- **Review**
+  - **ReviewId** (PK)
+  - UserId (FK → User)
+  - GameId (FK → Game)
+  - Rating *(0.0–5.0 in 0.5 steps, optional)*
+  - Text
+  - HasSpoilers
+  - CreatedAt
+  - UpdatedAt
+
+### Social
+- **Follow**
+  - **FollowId** (PK)
+  - FollowerId (FK → User)
+  - FollowingId (FK → User)
+  - CreatedAt
+  - **Constraint:** unique (FollowerId, FollowingId)
+
+- **ReviewLike**
+  - **ReviewLikeId** (PK)
+  - UserId (FK → User)
+  - ReviewId (FK → Review)
+  - CreatedAt
+  - **Constraint:** unique (UserId, ReviewId)
+
+- **ReviewComment**
+  - **ReviewCommentId** (PK)
+  - UserId (FK → User)
+  - ReviewId (FK → Review)
+  - Text
+  - CreatedAt
+
+- *(Optional stretch)* **List**
+  - **ListId** (PK)
+  - UserId (FK → User)
+  - Name
+  - Description *(optional)*
+  - CreatedAt
+  - UpdatedAt
+
+- *(Optional stretch)* **ListItem**
+  - **ListItemId** (PK)
+  - ListId (FK → List)
+  - GameId (FK → Game)
+  - Notes *(optional)*
+  - SortOrder *(optional)*
+  - CreatedAt
+  - **Constraint:** unique (ListId, GameId)
+
+---
+
+## API endpoints (draft)
+
+### Auth
+- `POST /api/auth/register`
+- `POST /api/auth/login`
+- `GET /api/auth/me`
+
+### Games + IGDB import
+- `GET /api/games` (search/filter/paging — local cache)
+- `GET /api/games/{gameId}`
+- `GET /api/igdb/search?query=...`
+- `POST /api/igdb/import/{igdbId}` *(admin-only or authenticated-only; TBD)*
+
+### Library
+- `GET /api/library/me`
+- `POST /api/library` (add/update log)
+- `DELETE /api/library/{gameId}`
+
+### Reviews + interactions
+- `POST /api/reviews`
+- `PUT /api/reviews/{reviewId}`
+- `DELETE /api/reviews/{reviewId}`
+- `POST /api/reviews/{reviewId}/like`
+- `DELETE /api/reviews/{reviewId}/like`
+- `POST /api/reviews/{reviewId}/comments`
+- `DELETE /api/comments/{reviewCommentId}` *(admin or owner)*
+
+### Social feed + follows
+- `GET /api/feed`
+- `POST /api/follows/{userId}`
+- `DELETE /api/follows/{userId}`
+
+### AI (core)
+- `POST /api/ai/recommendations`
+- `POST /api/ai/review-assistant`
+- `GET /api/ai/semantic-search?query=...`
+
+---
+
+## Technical requirements 
+
+- Front end + back end + database
+- AI is part of **core functionality** (recommendations)
+- Vector search is included (Azure AI Search)
+- Deployed to Azure with a CI/CD pipeline
+- Good architecture and coding practices:
+  - backend uses controllers/services/DTOs
+  - front end uses components/composables/services
+- Unit tests cover core functionality for front end and back end
+- UI/UX is polished, responsive, and enjoyable to use
+
+---
+
+## Testing plan
+
+### Front end (Vitest)
+- Store tests (Pinia): auth token persistence, role gating behavior
+- Service tests: games/library/reviews/feed API calls (axios mocked)
+- Component tests for critical flows:
+  - login/register form behavior
+  - add/update library status
+  - review create/edit + spoiler toggle
+  - feed renders + like/comment interactions
+
+### Back end (xUnit or NUnit)
+- Service-layer tests:
+  - library updates (status transitions, rating validation, ownership)
+  - review creation/edit rules + spoiler flag behavior
+  - follow rules (no self-follow, no duplicates)
+  - likes/comments rules (idempotency, ownership)
+- Controller tests:
+  - validation errors return correct status codes
+  - auth required for protected routes
+  - role-gated admin actions
+- Integration-style tests (if time allows):
+  - WebApplicationFactory for a few “happy path” endpoint flows
+
+---
+
+## Non-functional requirements
+
+- Performance: pagination on feeds and lists; avoid loading huge datasets
+- Security: JWT auth; role-based authorization; server-side validation
+- Reliability: friendly errors; consistent API responses; logging
+- Accessibility: readable contrast, keyboard navigation for key flows
+
+---
+
+## Milestones (suggested)
+
+- **M1:** Repo scaffold, CI/CD, auth (JWT), base UI shell
+- **M2:** IGDB search/import + game pages + library logging
+- **M3:** Reviews + follow + feed + likes/comments
+- **M4:** Azure AI Search + embeddings + recommendations + semantic search
+- **M5:** Writing assistant + polish + testing push + bug bash