Add personal access tokens, deployment management, storage profiles and remove MySQL

.coderabbit.yaml

@@ -0,0 +1,6 @@
+language: en-US
+early_access: false
+reviews:
+  path_filters:
+    - "!apps/backend/migrations/mysql/**"
+    - "!apps/backend/src/repository/mysql/**"

.github/workflows/ci.yml

@@ -134,10 +134,9 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        # SQLite runs fully parallel (in-memory, isolated per test). Postgres and
-        # MySQL run database-per-test against one shared server, so concurrent
-        # tests are capped to keep the connection demand well under the server's
-        # ceiling on high-core runners.
+        # SQLite runs fully parallel (in-memory, isolated per test). Postgres runs
+        # database-per-test against one shared server, so concurrent tests are
+        # capped to keep connection demand well under server ceiling.
         include:
           - db: sqlite
             test_database: sqlite
@@ -147,14 +146,10 @@ jobs:
             test_database: postgres
             test_database_url: postgres://postgres:postgres@localhost:5432/postgres
             test_args: "--test-threads 4"
-          - db: mysql
-            test_database: mysql
-            test_database_url: mysql://root:root@localhost:3306/mysql
-            test_args: "--test-threads 4"
     env:
       TEST_DATABASE: ${{ matrix.test_database }}
       TEST_DATABASE_URL: ${{ matrix.test_database_url }}
-    # Both services run for every lane; the sqlite lane simply ignores them
+    # Service runs for every lane; sqlite lane simply ignores it
     # (GitHub Actions can't attach services conditionally per matrix value).
     # Service containers are Linux-only, which is why Windows/macOS live in the
     # separate `test-os` job (SQLite only).
@@ -172,17 +167,6 @@ jobs:
           --health-interval 2s
           --health-timeout 5s
           --health-retries 15
-      mysql:
-        image: mysql:8
-        env:
-          MYSQL_ROOT_PASSWORD: root
-        ports:
-          - 3306:3306
-        options: >-
-          --health-cmd "mysqladmin ping -h 127.0.0.1 -uroot -proot"
-          --health-interval 2s
-          --health-timeout 5s
-          --health-retries 15
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
       - uses: arduino/setup-protoc@c65c819552d16ad3c9b72d9dfd5ba5237b9c906b # v3
@@ -195,7 +179,7 @@ jobs:
         with:
           # Workspace root is the REPO root (Cargo.lock + target/ live there).
           workspaces: ". -> target"
-          # All 3 db lanes compile an identical binary (db is chosen at runtime
+          # Both db lanes compile an identical binary (db is chosen at runtime
           # via TEST_DATABASE), so one shared cache serves all of them; the
           # save race between lanes is harmless. NOT shared with lint: clippy
           # saves check-artifacts that are useless for test codegen.

AGENTS.md

@@ -8,7 +8,7 @@
 
 ## Architecture
 
-- **Backend**: Rust + Axum HTTP server, `SQLx` + (SQLite | PostgreSQL | MySQL), `rust-embed` (static assets)
+- **Backend**: Rust + Axum HTTP server, `SQLx` + (SQLite | PostgreSQL), `rust-embed` (static assets)
 - **Frontend**: React in `apps/dashboard/` with Tanstack Router, Tanstack Query, shadcn/ui
 - **gRPC**: Separate server on port 50051 (compiled from `libs/proto/*.proto` via `tonic-build`)
 - **Build**: Nx orchestrates `dashboard:build` → `backend:build`. `build.rs` compiles proto files only.
@@ -222,7 +222,7 @@ A logical backup/restore subsystem lives in `apps/backend/src/services/backup/`:
 - `schema.rs` — table registry + cross-backend dump/restore. Every value is
   normalized to **text** (Postgres casts `::text`/`::int`; restore casts back with
   `::jsonb`/`::timestamptz`/`::bigint`/`::int::boolean`), so a backup is a portable,
-  DB-agnostic set of NDJSON rows that restores into any of the three backends.
+  DB-agnostic set of NDJSON rows that restores into either backend.
 - `mod.rs` — `BackupService`: a snapshot read (REPEATABLE READ / WAL) dumps tables →
   tar → zstd → optional AES-256-GCM, written to a destination `StorageProvider`.
   Restore is full-replace within the chosen scope in one transaction (site filter,

CLAUDE.md

@@ -2,7 +2,7 @@
 
 ## Architecture
 
-- **Backend**: Rust + Axum HTTP server, `SQLx` + (SQLite | PostgreSQL | MySQL), `rust-embed` (static assets)
+- **Backend**: Rust + Axum HTTP server, `SQLx` + (SQLite | PostgreSQL), `rust-embed` (static assets)
 - **Frontend**: React in `apps/dashboard/` with Tanstack Router, Tanstack Query, shadcn/ui
 - **gRPC**: Separate server on port 50051 (compiled from `libs/proto/*.proto` via `tonic-build`)
 - **Build**: Nx orchestrates `dashboard:build` → `backend:build`. `build.rs` compiles proto files only.
@@ -225,8 +225,8 @@ a backup taken on a **newer** schema and relies on additive migrations otherwise
 
 Entry search is backed by an embedded [Tantivy](https://github.com/quickwit-oss/tantivy)
 index in `apps/backend/src/services/search/` — one engine that gives ranked,
-tokenized, stemmed search identically across SQLite/Postgres/MySQL (native per-DB
-FTS would mean three implementations). The DB stays the source of truth; the index
+tokenized, stemmed search identically across SQLite/Postgres (native per-DB
+FTS would mean two implementations). The DB stays the source of truth; the index
 is **derived** and fully rebuildable.
 
 **Reads vs writes are split** so the single-writer limit never blocks *searching*:

README.md

@@ -49,7 +49,7 @@ Upload, organize, and serve images and files with automatic thumbnail generation
 
 ### 💾 Backups & Disaster Recovery
 
-Create on-demand or scheduled backups of a single site or the whole instance. Backups are compressed, optionally encrypted (AES-256-GCM), and stored on local disk or a separate S3 bucket — keep the last N automatically. Restore in place or import a site as a copy, from the dashboard or offline with `vcms restore` when the server won't even boot. Backups are a portable logical dump, so you can move data between SQLite, PostgreSQL, and MySQL.
+Create on-demand or scheduled backups of a single site or the whole instance. Backups are compressed, optionally encrypted (AES-256-GCM), and stored on local disk or a separate S3 bucket — keep the last N automatically. Restore in place or import a site as a copy, from the dashboard or offline with `vcms restore` when the server won't even boot. Backups are a portable logical dump, so you can move data between SQLite and PostgreSQL.
 
 ### 🔐 Secure & Scalable
 
@@ -103,10 +103,11 @@ that the client process can't read.
 VCMS_MCP_TOKEN=vcms_site_... VCMS_MCP_URL=http://127.0.0.1:3000 vcms mcp stdio
 ```
 
-It needs only two env vars: `VCMS_MCP_TOKEN` (a `vcms_site_*` access token, forwarded
-as the `Authorization: Bearer` credential) and `VCMS_MCP_URL` (the running server's
-base URL, default `http://127.0.0.1:3000`; the proxy posts to `{url}/mcp`). A `vcms
-serve` instance must be running. MCP protocol messages use stdout; logs use stderr.
+It needs only two env vars: `VCMS_MCP_TOKEN` (a `vcms_site_*` or `vcms_pat_*`
+access token with the `mcp.use` scope, forwarded as the `Authorization: Bearer`
+credential) and `VCMS_MCP_URL` (the running server's base URL, default
+`http://127.0.0.1:3000`; the proxy posts to `{url}/mcp`). A `vcms serve` instance
+must be running. MCP protocol messages use stdout; logs use stderr.
 
 ```jsonc
 // Example MCP client config
@@ -166,7 +167,7 @@ Built by developers, for developers. The API is predictable, the documentation i
 
 ### Database Flexibility
 
-Use SQLite for simple deployments or connect to PostgreSQL or MySQL for production workloads. The same binary works with all three.
+Use SQLite for simple deployments or connect to PostgreSQL for production workloads. The same binary works with both.
 
 ### Built for Teams
 

apps/backend/Cargo.toml

@@ -28,7 +28,7 @@ tower = { version = "0.5", features = ["full"] }
 tower-http = { version = "0.7", features = ["cors", "limit", "trace"] }
 rust-embed = { version = "8", optional = true }
 mime_guess = "2"
-sqlx = { version = "0.9", features = ["runtime-tokio", "macros", "postgres", "mysql", "sqlite", "chrono", "uuid", "migrate"] }
+sqlx = { version = "0.9", features = ["runtime-tokio", "macros", "postgres", "sqlite", "chrono", "uuid", "migrate"] }
 thiserror = "2"
 async-trait = "0.1"
 uuid = { version = "1", features = ["v4", "v7", "serde"] }