Understand impact, lineage, and SQL dependencies faster. Build interactive maps from DACPACs or live Microsoft SQL metadata without leaving VS Code.
Free core experience · MIT Licensed · Optional Copilot AI
Untangle long SQL chains, inherited models, and undocumented reporting logic without manual tracing.
Check downstream impact before you change a table, view, or reporting calculation.
Stay in VS Code with a local-first graph that makes architectural intent easier to inspect and explain.
Move from question to impact analysis without leaving VS Code.
Spot risky dependencies, hidden hubs, and isolated objects before a change reaches production.
Keep the SQL beside the graph so you can verify logic faster and document decisions while you work.
Start large estates in schema view so the graph stays readable and responsive.
Run safe aggregate checks on live tables when you need a quick sanity check before deeper analysis.
Save views, keep filters, and revisit important paths without rebuilding context.
Adjust parse and AI behavior with workspace settings and YAML-based templates.
Local extraction, graph traversal, and optional AI guidance designed for large SQL estates.
Securely extract metadata from local .dacpac files offline, or connect to live MS SQL projects via standard DMVs.
A staged parser extracts references from T-SQL bodies while filtering strings, comments, and noisy syntax patterns.
Graphology plus deterministic BFS powers fast dependency traversal, scope discovery, and path finding.
When AI is enabled, the assistant works from bounded metadata summaries so longer traces stay explainable without losing control.
When GitHub Copilot is enabled, @lineage answers business or technical questions from the loaded model and sends you back into a graph you can keep exploring.
Ask direct questions about formulas, dependencies, impact, and reporting logic.
The assistant works from your loaded metadata, not from generic SQL guesses.
Keep exploring the returned graph with search, trace tools, and saved views inside the same workflow.
A representative AI result inside the graph: linked objects, summarized findings, and visual context you can keep exploring after the answer is generated.
Recorded example prompt:
Requires GitHub Copilot & VS Code 1.95+ · Core graph exploration remains local and available without AI
No. DDL extraction, graph rendering, and core lineage analysis stay local. If you enable the optional AI features, only the metadata needed for the current answer is sent to your configured GitHub Copilot endpoint.
Lineage tracing and AI analysis use compile-time metadata and DDL. Row-level data is accessed only if you manually trigger Table Profiling, and no row-level data is sent to the AI.
The engine is natively optimized for Microsoft SQL Server, Azure SQL Database, Microsoft Fabric DW, and Azure Synapse Analytics.
Architectures exceeding your configured threshold automatically trigger "Schema Overview" mode, collapsing objects into manageable schema bubbles to prevent UI lag. For deep dependency chains, the AI utilizes a bounded "Sliding Memory" window to process traces without context exhaustion.
Yes. The extraction engine is fully metadata-driven. You can override T-SQL regex parsing rules, profiling statistics patterns, and AI generation prompts using local YAML templates within your workspace.
Built by Christian Wagner, a data warehousing and engineering practitioner working across SQL Server, Synapse, Fabric, and Azure-based analytics platforms.
The goal is simple: make SQL lineage easier to inspect, explain, and improve inside the tools teams already use. Feedback, bug reports, and practical product ideas are always welcome.
Connect on LinkedIn
Install Extension