Beerust 1.2.0

Added

• bee_orm: the ORM now runs end to end — typed Value bind parameters (Null / Bool / Int / Float / Text / Bytes), #[derive(Model)] with #[bee(table / column / pk / auto / ignore)], model insert / update / delete, and QuerySet execution (all / one / count / exists / update / delete)
• bee_orm: connection pools for all three backends (pool::{sqlite, postgres, mysql}::Pool) — connect(dsn, max_size), get() → CheckedConn, query / execute, status(), and transactions via begin / commit / rollback; a connection dropped mid-transaction is rolled back on sqlite and postgres
• bee_orm: new postgres-tls feature (PostgreSQL over TLS with bundled Mozilla roots) and pool hardening — 30 s wait / 10 s create timeouts plus a prepared-statement cache on the postgres pool
• bee_orm: model lifecycle — #[bee(auto_now_add)] / #[bee(auto_now)] timestamps, #[bee(soft_delete)] (delete flips the flag; with_deleted() / hard_delete() escape it), lifecycle hooks (before/after insert/update/delete), Model::insert_many (999-parameter chunks, not transactional) and QuerySet::filter_in plus sum / avg / min / max aggregates
• bee_orm: non-destructive migrations — migrate::{create_table, add_missing_columns, sync} render dialect-aware DDL (sqlite AUTOINCREMENT / postgres IDENTITY / mysql AUTO_INCREMENT), adding tables and missing columns but never dropping or altering
• bee_orm: foreign-key relations — #[bee(fk = Target)] (with #[bee(sql_type = "…")]) emits FK DDL, and rel::{fk_column_to, belongs_to, children, children_for} read them; enforcement follows the backend: postgres and sqlite (bundled build) enforce the reference, mysql ignores inline REFERENCES (a gap for round 5+)
• bee_orm: many-to-many relations — #[bee(m2m(Target))] declares the relation, m2m::{attach, detach, related, related_for, related_ids} read and write it, and create_table / sync create the join table (add_missing_columns never touches it)
• bee_orm: JSON columns — serde_json::Value fields map to TEXT / JSONB / JSON per backend, with honest NULL semantics: SQL NULL decodes to None, a stored JSON null document to Some(Json::Null)
• bee_orm: opt-in MySQL table-level foreign keys — MigrateOptions { table_level_fk } with sync_with / create_table_with / add_missing_columns_with; off by default, and pre-existing orphan rows make ADD CONSTRAINT fail instead of being skipped silently
• bee_orm: Pool::connect_tls_with for a custom rustls::ClientConfig, plus a bee_orm::rustls re-export so the config version always matches; connect_tls keeps the bundled webpki roots
• bee_rust: four forwarding features — orm-sqlite / orm-postgres / orm-postgres-tls / orm-mysql forward a bee_orm backend through bee_rust (none of them are in full)
• bee_cli: bee-rust migrate init scaffolds src/bin/bee_migrate.rs (refuses to overwrite an existing file) and bee-rust migrate run executes it via cargo run --bin bee_migrate
• bee_kv / bee_cache: the Redis backends survive dropped connections — the underlying ConnectionManager reconnects on demand (no background thread) with exponential backoff and jitter (the command that hits the break errors; the next one waits for the fresh connection)
• bee_kv / bee_cache: memcached backends (MemcacheStore in bee_kv, MemcacheCache in bee_cache); incr creates the counter before applying the delta, unsigned counters floor at 0, and TTL 0 deletes the key

Changed

• bee_orm migrations, relations (including many-to-many), the bee-rust migrate CLI subcommand, the Redis / Memcached backends, and the bee_rust ORM forwarding features are no longer marked as planned in the docs
• the search / graph / tsdb drivers are no longer marked in the docs as planned or as a trait stub, but as implemented (opt-in features); the old examples were corrected (nonexistent ElasticsearchEngine / Neo4jDB type names, wrong bolt:// scheme)