Cross-Database Similarity & Migration
Q4 2026Deterministic + AI scoring identifies duplicate objects across warehouses; generates migration plans with column mappings and reconciliation queries.
Quarter-granular view of what BoltPipeline ships, what's in flight, what's next, and what we're considering. No exact dates, no marketing fluff — derived from the engineering spec backlog.
Already generally available. Each shipped capability has its own deep-dive page with what it solves, how it works, and what it connects to.
Browse all 27 shipped features →Actively being built. Expect these to land within the current quarter.
Deterministic + AI scoring identifies duplicate objects across warehouses; generates migration plans with column mappings and reconciliation queries.
Closes a high-priority gap where agent OPERATE compile-failure-all-steps silently drops the result publish, leaving pipelines stuck.
Drafted and queued. Sequencing is set by the engineering backlog.
Zero-config, metadata-first DQ: composite health score per table, freshness via Snowflake metadata, seasonal volume baseline, column-level anomaly, lineage-aware RCA.
Per-table freshness SLA hours + 15-minute fleet-wide evaluator; surfaced on Monitor Live and configurable from Settings.
Per-table volume drift threshold with 7-snapshot rolling baseline; sparkline on Monitor; tenant-default + per-table override.
Per-tenant z-score thresholds over baseline view; v1 covers null_pct, distinct_count, min/max; heat-tile + drawer on Monitor Assets.
Native glossary terms + links into Catalog with Table-360 badges and command-K search extension.
5-state lifecycle (draft → proposed → under_review → approved → deprecated) on top of the glossary, with a steward role and reuse of existing workflow engine.
Tag taxonomy + per-column tags with downstream propagation through column lineage; shield badges on Lineage and Table-360 surfaces.
Import an existing dbt project through one façade SP that fans out to pipeline write, asset graph, and cert rules — same governance, no rewrite.
Pulls Unity Catalog precomputed table + column lineage into BoltPipeline's vendor-neutral lineage graph; [DBX] vendor prefix in the DAG.
Daily cost rollup keyed by pipeline_guid in QUERY_TAG; new Operate → Pipelines → Cost tab; extends services/cost_analyzer in place.
Recursive upstream walk over column lineage joined with drift events in a temporal window; deterministic ranking surfaces the most-likely root drift.
New agent vendor framework + replication-pair model; cross-vendor Catalog DAG and pair-detail diff UI for SQL Server replication health.
Extends the pipeline watchdog with an agent-dead-no-publish branch so stuck tasks fail fast instead of waiting the full 45-minute worst case.
Lifts the activity-summary computation out of the midnight cron behind a rate-limited refresh op so operators can pull current-day rollups on demand.
Ideas we believe in but have not committed to. No target date is implied.
Persist agent run state across cold restarts so long-running pipelines survive infra churn without operator nudging.
Allow the agent to keep certifying and queueing work for a bounded window when the control plane is unreachable.
Sub-tenant administrative scope so a single tenant can delegate per-department admin without granting full tenant-admin.
Conversational assistant grounded in the tenant's catalog + lineage to answer impact, ownership, and drift questions.
Optionally project BoltPipeline lineage into OpenLineage / Datakin schemas for ecosystem interoperability.
Richer artifact versioning model with branch-aware semantics and reproducible promotion across environments.
First-class console editor for logical → physical FQN aliases (today configured by API + DB).
One user can belong to multiple teams within a tenant with team-scoped roles and visibility.
Single user with multiple tenant memberships can switch context without re-authenticating.
Richer per-run execution manifest with embedded artifact references for forensic replay and audit.
Operator-driven requeue of stuck tasks with edited context (e.g., updated step inputs) instead of brute-force restart.
Customer signal drives sequencing. Tell us what would unblock you and we will weight it on the next backlog review.