Add multi-database data layer

Introduce aether-data-schema and driver-specific schema generation for Postgres, MySQL, and SQLite.

Split data backends, lifecycle, repositories, and gateway runtime integration across database drivers.

Verified with cargo fmt --all --check, cargo clippy --workspace --all-targets -- -D warnings, and cargo test --workspace.

crates/aether-data-contracts/Cargo.toml

@@ -9,6 +9,7 @@ description = "Shared data contracts and repository traits for Aether Rust servi
 [dependencies]
 aether-ai-formats.workspace = true
 async-trait.workspace = true
+chrono.workspace = true
 serde.workspace = true
 serde_json.workspace = true
 thiserror.workspace = true

crates/aether-data-contracts/src/error.rs

@@ -12,6 +12,9 @@ pub enum DataLayerError {
     #[error("redis error: {0}")]
     Redis(String),
 
+    #[error("sql error: {0}")]
+    Sql(String),
+
     #[error("operation timed out: {0}")]
     TimedOut(String),
 
@@ -27,4 +30,8 @@ impl DataLayerError {
     pub fn redis(error: impl std::fmt::Display) -> Self {
         Self::Redis(error.to_string())
     }
+
+    pub fn sql(error: impl std::fmt::Display) -> Self {
+        Self::Sql(error.to_string())
+    }
 }

crates/aether-data-contracts/src/repository/billing/mod.rs

@@ -1,7 +1,7 @@
 mod types;
 
 pub use types::{
-    AdminBillingCollectorRecord, AdminBillingCollectorWriteInput, AdminBillingPresetApplyResult,
-    AdminBillingRuleRecord, AdminBillingRuleWriteInput, BillingReadRepository,
-    StoredBillingModelContext,
+    AdminBillingCollectorRecord, AdminBillingCollectorWriteInput, AdminBillingMutationOutcome,
+    AdminBillingPresetApplyResult, AdminBillingRuleRecord, AdminBillingRuleWriteInput,
+    BillingReadRepository, StoredBillingModelContext,
 };

crates/aether-data-contracts/src/repository/billing/types.rs

@@ -142,6 +142,14 @@ pub struct AdminBillingPresetApplyResult {
     pub errors: Vec<String>,
 }
 
+#[derive(Debug, Clone, PartialEq)]
+pub enum AdminBillingMutationOutcome<T> {
+    Applied(T),
+    NotFound,
+    Invalid(String),
+    Unavailable,
+}
+
 #[async_trait]
 pub trait BillingReadRepository: Send + Sync {
     async fn find_model_context(
@@ -160,4 +168,109 @@ pub trait BillingReadRepository: Send + Sync {
         let _ = (provider_id, provider_api_key_id, model_id);
         Ok(None)
     }
+
+    async fn admin_billing_enabled_default_value_exists(
+        &self,
+        api_format: &str,
+        task_type: &str,
+        dimension_name: &str,
+        existing_id: Option<&str>,
+    ) -> Result<Option<bool>, crate::DataLayerError> {
+        let _ = (api_format, task_type, dimension_name, existing_id);
+        Ok(None)
+    }
+
+    async fn create_admin_billing_rule(
+        &self,
+        input: &AdminBillingRuleWriteInput,
+    ) -> Result<AdminBillingMutationOutcome<AdminBillingRuleRecord>, crate::DataLayerError> {
+        let _ = input;
+        Ok(AdminBillingMutationOutcome::Unavailable)
+    }
+
+    async fn list_admin_billing_rules(
+        &self,
+        task_type: Option<&str>,
+        is_enabled: Option<bool>,
+        page: u32,
+        page_size: u32,
+    ) -> Result<Option<(Vec<AdminBillingRuleRecord>, u64)>, crate::DataLayerError> {
+        let _ = (task_type, is_enabled, page, page_size);
+        Ok(None)
+    }
+
+    async fn find_admin_billing_rule(
+        &self,
+        rule_id: &str,
+    ) -> Result<Option<AdminBillingRuleRecord>, crate::DataLayerError> {
+        let _ = rule_id;
+        Ok(None)
+    }
+
+    async fn update_admin_billing_rule(
+        &self,
+        rule_id: &str,
+        input: &AdminBillingRuleWriteInput,
+    ) -> Result<AdminBillingMutationOutcome<AdminBillingRuleRecord>, crate::DataLayerError> {
+        let _ = (rule_id, input);
+        Ok(AdminBillingMutationOutcome::Unavailable)
+    }
+
+    async fn create_admin_billing_collector(
+        &self,
+        input: &AdminBillingCollectorWriteInput,
+    ) -> Result<AdminBillingMutationOutcome<AdminBillingCollectorRecord>, crate::DataLayerError>
+    {
+        let _ = input;
+        Ok(AdminBillingMutationOutcome::Unavailable)
+    }
+
+    async fn list_admin_billing_collectors(
+        &self,
+        api_format: Option<&str>,
+        task_type: Option<&str>,
+        dimension_name: Option<&str>,
+        is_enabled: Option<bool>,
+        page: u32,
+        page_size: u32,
+    ) -> Result<Option<(Vec<AdminBillingCollectorRecord>, u64)>, crate::DataLayerError> {
+        let _ = (
+            api_format,
+            task_type,
+            dimension_name,
+            is_enabled,
+            page,
+            page_size,
+        );
+        Ok(None)
+    }
+
+    async fn find_admin_billing_collector(
+        &self,
+        collector_id: &str,
+    ) -> Result<Option<AdminBillingCollectorRecord>, crate::DataLayerError> {
+        let _ = collector_id;
+        Ok(None)
+    }
+
+    async fn update_admin_billing_collector(
+        &self,
+        collector_id: &str,
+        input: &AdminBillingCollectorWriteInput,
+    ) -> Result<AdminBillingMutationOutcome<AdminBillingCollectorRecord>, crate::DataLayerError>
+    {
+        let _ = (collector_id, input);
+        Ok(AdminBillingMutationOutcome::Unavailable)
+    }
+
+    async fn apply_admin_billing_preset(
+        &self,
+        preset: &str,
+        mode: &str,
+        collectors: &[AdminBillingCollectorWriteInput],
+    ) -> Result<AdminBillingMutationOutcome<AdminBillingPresetApplyResult>, crate::DataLayerError>
+    {
+        let _ = (preset, mode, collectors);
+        Ok(AdminBillingMutationOutcome::Unavailable)
+    }
 }

crates/aether-data-contracts/src/repository/usage/mod.rs

@@ -1,21 +1,23 @@
 mod types;
 
 pub use types::{
-    parse_usage_body_ref, usage_body_ref, StoredProviderApiKeyUsageSummary,
-    StoredProviderUsageSummary, StoredProviderUsageWindow, StoredRequestUsageAudit,
-    StoredUsageAuditAggregation, StoredUsageAuditSummary, StoredUsageBreakdownSummaryRow,
-    StoredUsageCacheAffinityHitSummary, StoredUsageCacheAffinityIntervalRow,
-    StoredUsageCacheHitSummary, StoredUsageCostSavingsSummary, StoredUsageDailySummary,
-    StoredUsageDashboardDailyBreakdownRow, StoredUsageDashboardProviderCount,
-    StoredUsageDashboardSummary, StoredUsageErrorDistributionRow, StoredUsageLeaderboardSummary,
+    parse_usage_body_ref, usage_body_ref, PendingUsageCleanupSummary,
+    StoredProviderApiKeyUsageSummary, StoredProviderUsageSummary, StoredProviderUsageWindow,
+    StoredRequestUsageAudit, StoredUsageAuditAggregation, StoredUsageAuditSummary,
+    StoredUsageBreakdownSummaryRow, StoredUsageCacheAffinityHitSummary,
+    StoredUsageCacheAffinityIntervalRow, StoredUsageCacheHitSummary, StoredUsageCostSavingsSummary,
+    StoredUsageDailySummary, StoredUsageDashboardDailyBreakdownRow,
+    StoredUsageDashboardProviderCount, StoredUsageDashboardSummary,
+    StoredUsageErrorDistributionRow, StoredUsageLeaderboardSummary,
     StoredUsagePerformancePercentilesRow, StoredUsageSettledCostSummary,
     StoredUsageTimeSeriesBucket, StoredUsageUserTotals, UpsertUsageRecord,
     UsageAuditAggregationGroupBy, UsageAuditAggregationQuery, UsageAuditKeywordSearchQuery,
     UsageAuditListQuery, UsageAuditSummaryQuery, UsageBodyCaptureResult, UsageBodyCaptureState,
     UsageBodyCaptureStorage, UsageBodyField, UsageBreakdownGroupBy, UsageBreakdownSummaryQuery,
     UsageCacheAffinityHitSummaryQuery, UsageCacheAffinityIntervalGroupBy,
-    UsageCacheAffinityIntervalQuery, UsageCacheHitSummaryQuery, UsageCostSavingsSummaryQuery,
-    UsageDailyHeatmapQuery, UsageDashboardDailyBreakdownQuery, UsageDashboardProviderCountsQuery,
+    UsageCacheAffinityIntervalQuery, UsageCacheHitSummaryQuery, UsageCleanupSummary,
+    UsageCleanupWindow, UsageCostSavingsSummaryQuery, UsageDailyHeatmapQuery,
+    UsageDashboardDailyBreakdownQuery, UsageDashboardProviderCountsQuery,
     UsageDashboardSummaryQuery, UsageErrorDistributionQuery, UsageLeaderboardGroupBy,
     UsageLeaderboardQuery, UsageMonitoringErrorCountQuery, UsageMonitoringErrorListQuery,
     UsagePerformancePercentilesQuery, UsageReadRepository, UsageRepository,

crates/aether-data-contracts/src/repository/usage/types.rs

@@ -1,4 +1,5 @@
 use async_trait::async_trait;
+use chrono::{DateTime, Utc};
 use serde_json::Value;
 
 /// Joined usage read model assembled from the accounting row plus the newer audit/snapshot
@@ -1568,12 +1569,57 @@ pub trait UsageWriteRepository: Send + Sync {
     async fn rebuild_api_key_usage_stats(&self) -> Result<u64, crate::DataLayerError>;
 
     async fn rebuild_provider_api_key_usage_stats(&self) -> Result<u64, crate::DataLayerError>;
+
+    async fn cleanup_stale_pending_requests(
+        &self,
+        cutoff_unix_secs: u64,
+        now_unix_secs: u64,
+        timeout_minutes: u64,
+        batch_size: usize,
+    ) -> Result<PendingUsageCleanupSummary, crate::DataLayerError> {
+        let _ = (cutoff_unix_secs, now_unix_secs, timeout_minutes, batch_size);
+        Ok(PendingUsageCleanupSummary::default())
+    }
+
+    async fn cleanup_usage(
+        &self,
+        window: &UsageCleanupWindow,
+        batch_size: usize,
+        auto_delete_expired_keys: bool,
+    ) -> Result<UsageCleanupSummary, crate::DataLayerError> {
+        let _ = (window, batch_size, auto_delete_expired_keys);
+        Ok(UsageCleanupSummary::default())
+    }
 }
 
 pub trait UsageRepository: UsageReadRepository + UsageWriteRepository + Send + Sync {}
 
 impl<T> UsageRepository for T where T: UsageReadRepository + UsageWriteRepository + Send + Sync {}
 
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
+pub struct PendingUsageCleanupSummary {
+    pub failed: usize,
+    pub recovered: usize,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default, serde::Serialize, serde::Deserialize)]
+pub struct UsageCleanupSummary {
+    pub body_externalized: usize,
+    pub legacy_body_refs_migrated: usize,
+    pub body_cleaned: usize,
+    pub header_cleaned: usize,
+    pub keys_cleaned: usize,
+    pub records_deleted: usize,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
+pub struct UsageCleanupWindow {
+    pub detail_cutoff: DateTime<Utc>,
+    pub compressed_cutoff: DateTime<Utc>,
+    pub header_cutoff: DateTime<Utc>,
+    pub log_cutoff: DateTime<Utc>,
+}
+
 fn parse_u64(value: i32, field_name: &str) -> Result<u64, crate::DataLayerError> {
     u64::try_from(value).map_err(|_| {
         crate::DataLayerError::UnexpectedValue(format!("invalid {field_name}: {value}"))