fix(storage): use current tenant for account snapshots

[gouhongshen]

Summary

• remove the hardcoded FOR ACCOUNT sys from execute-path pre-migration account snapshots
• align legacy single-db safety snapshots to use the same current-tenant account snapshot form
• update the git-for-data RFC examples to match the tenant-safe MatrixOne syntax

Validation

• cargo check -p memoria-storage -p memoria-cli
• real memoria migrate legacy-to-multi-db --execute on a non-sys tenant test DB
• inspected migrate-run.json: no warnings, no row mismatches
• SQL spot-check: sampled mem_user_registry mapping and mem_memories counts match legacy/user DB

[Copilot]

Pull request overview

This PR updates snapshot/migration behavior to be tenant-safe and adds multi-database (“shared DB + per-user DB”) routing support across the storage layer, service layer, CLI, MCP tools, and API endpoints.

Changes:

• Remove hardcoded FOR ACCOUNT sys snapshot usage and align snapshot tooling to current-tenant / DB-scoped semantics.
• Introduce DbRouter + legacy single-DB → multi-DB migration utilities, and wire routing through service/API/MCP paths.
• Expand/adjust integration tests and RFC documentation examples to reflect the new snapshot + routing behavior.

Reviewed changes

Copilot reviewed 35 out of 36 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
memoria/docs/git-for-data-rfc.md	Updates snapshot SQL examples (but still needs alignment with DB-scoped implementation).
memoria/crates/memoria-test-utils/src/lib.rs	Adds MySQL readiness helper for integration tests.
memoria/crates/memoria-test-utils/Cargo.toml	Adds sqlx dependency for readiness helper.
memoria/crates/memoria-storage/tests/router_multi_db.rs	Adds integration test validating user isolation across routed DBs.
memoria/crates/memoria-storage/src/router.rs	Introduces DbRouter for user→DB routing with caches and provisioning.
memoria/crates/memoria-storage/src/migration.rs	Adds legacy single-DB → multi-DB migration planner/executor with snapshots.
memoria/crates/memoria-storage/src/lib.rs	Exports new migration/router APIs.
memoria/crates/memoria-storage/Cargo.toml	Adds sha2 dependency for deterministic user DB naming.
memoria/crates/memoria-service/tests/plugin_repository.rs	Extends service config used in tests for new fields.
memoria/crates/memoria-service/src/service.rs	Routes per-user operations via DbRouter; adjusts access counter + retrieval param lookups.
memoria/crates/memoria-service/src/scheduler.rs	Updates test config initialization for new config fields.
memoria/crates/memoria-service/src/pipeline.rs	Makes persist phase resolve the correct per-user SQL store.
memoria/crates/memoria-service/src/governance.rs	Executes governance operations across routed per-user stores when enabled.
memoria/crates/memoria-service/src/config.rs	Adds shared_db_url + multi_db config and helpers.
memoria/crates/memoria-mcp/tests/snapshot_e2e.rs	Updates snapshot tests for scoped naming + adds concurrency/multi-db rollback tests.
memoria/crates/memoria-mcp/tests/edit_log_e2e.rs	Updates safety snapshot naming expectations to include DB scope.
memoria/crates/memoria-mcp/tests/branch_e2e.rs	Adds merge/replace regression test and test embedder wiring.
memoria/crates/memoria-mcp/src/tools.rs	Routes tool operations through per-user SQL stores.
memoria/crates/memoria-mcp/src/git_tools.rs	Scopes snapshot names by DB, adds duplicate-create coalescing, and improves safety snapshot visibility.
memoria/crates/memoria-git/src/service.rs	Creates DB-scoped snapshots and filters snapshot listing by database.
memoria/crates/memoria-cli/src/main.rs	Adds memoria migrate legacy-to-multi-db command; wires multi-db server startup + URL redaction.
memoria/crates/memoria-api/tests/pool_isolation.rs	Fixes db name parsing to ignore URL suffixes.
memoria/crates/memoria-api/tests/api_e2e.rs	Strengthens snapshot detail/diff assertions and adds structured branch list test.
memoria/crates/memoria-api/tests/api_db_verify.rs	Uses DB readiness helper for more reliable CI startup.
memoria/crates/memoria-api/src/state.rs	Routes background flushers through MemoryService for correct routing in multi-db mode.
memoria/crates/memoria-api/src/routes/snapshots.rs	Enhances snapshot endpoints with structured JSON and multi-db-aware snapshot resolution.
memoria/crates/memoria-api/src/routes/sessions.rs	Makes session summary reads branch-aware by using the active table.
memoria/crates/memoria-api/src/routes/metrics.rs	Makes metrics collection multi-db aware by aggregating per-user DBs.
memoria/crates/memoria-api/src/routes/memory.rs	Uses per-user SQL store for profile/history/retrieval param routes.
memoria/crates/memoria-api/src/routes/governance.rs	Uses per-user SQL store and makes cooldown keys user-scoped.
memoria/crates/memoria-api/src/routes/auth.rs	Uses shared/auth pool selection (no longer runs per-route DDL).
memoria/crates/memoria-api/src/routes/admin.rs	Updates admin stats and per-user operations for multi-db routing.
memoria/crates/memoria-api/src/auth.rs	Routes tool usage + call log batch flushes via MemoryService (per-user in multi-db).
memoria/Cargo.lock	Locks new dependencies (sha2, sqlx in test utils).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The RFC still documents snapshots as account-level (CREATE SNAPSHOT <name> FOR ACCOUNT / RESTORE ACCOUNT ...), but the implementation now creates DB-scoped snapshots (CREATE SNAPSHOT <name> FOR DATABASE <db>) and SHOW SNAPSHOTS results are filtered to the current database. This doc section should be updated to avoid suggesting account-level snapshot/restore semantics that won’t match the current Rust implementation.

[Copilot]

Snapshot API reads are now based on the user’s active branch table (let table = sql.active_table(...)). This can make /v1/snapshots/:name and /v1/snapshots/:name/diff return branch-table snapshot contents when the user is checked out to a branch, which is inconsistent with rollback behavior (rollback always restores mem_memories). Consider hardcoding these snapshot detail/diff queries to mem_memories (or otherwise explicitly define the intended semantics) so snapshot reads don’t depend on the current checkout state.

Suggested change

	let table = sql.active_table(&user_id).await.map_err(api_err)?;
	let table = "mem_memories";

[Copilot]

create_snapshot() relies on validate_identifier() which currently accepts any Unicode alphanumeric characters (char::is_alphanumeric). Elsewhere (e.g., memoria-mcp git tools) identifier validation is ASCII-only, and MatrixOne identifiers are typically ASCII-centric. Tightening validation to is_ascii_alphanumeric() (plus underscore) would make behavior consistent and avoid runtime DDL failures for identifiers that pass validation but are rejected by the database.

[gouhongshen]

Superseded by #169 to avoid conflicts from the previously merged feature-branch history.