Skip to content
Graphicient Wiki

Search is only available in production builds. Try building and previewing the site to test it out locally.

Release notes – Production deployment of 2026-10-12

🛠️ GraphForge

  • New Expert mode, which lets you:
    • Data tab:
      • Snowflake, Databricks and PostgreSQL connectors, with a source catalog (files or SQL).
      • SQL exploration console within connectors to test a query before adding it.
    • Transformation tab: Python code to apply to source data to transform it.
    • Mapping tab:
      • Guided mapping screen: columns → nodes and relationships, then loading into the graph.
      • Legacy mappings automatically migrated: no more empty catalogs.
    • Lineage Overview tab: see which data source has been mapped to which object/relationship in the ontology.
    • Pipeline Loads tab: pipelines of operations performed on data sources, with real-time monitoring and automatic recovery in case of incidents.
    • History tab (within the Ontology tab): ontology versioning.
  • Basic mode, which does everything you could do before, with a more user-friendly experience.
  • Additional new features:
    • Ontology quality KPIs.
    • Semantic tab: enrichment of the semantic layer and interaction with the AI features.
    • Semantic layer added globally for objects and relationships, and locally for attributes (long description, business rules and synonyms).
    • Confidentiality support: C1, C2, C3 and PII classifications now handled, data encryption, and access rights managed by application service (same as Snowflake).
    • AI Agent tab: interaction with the AI assistant, with the same features available.

🧭 Explore

  • New UI.
  • Advanced search bar: multi-attribute filters.
  • Comments: improved Comments feature.
  • Graph search added to the Visualizations feature.
  • Shortest path from an object: new “excluded type” option.
  • New Data > Breakdown tab: mini KPI plots on an object’s attributes.
  • Better interaction between Explore and AI.

🤖 AI

  • More autonomous Skill Creator: the AI drafts, corrects and reviews the skill on its own; saving is no longer blocked by formatting issues, only by a write operation on the graph or a reference to a non-existent schema element.
  • Skills with Python scripts: a skill can now embed its own calculation logic (reference .py file), validated on save.
  • Code Interpreter: Kepler can now run calculations (rates, reconciliations, pivots) on query results, within a secure internal execution environment.
  • New ontology (V2): the agent reads tags, synonyms, descriptions and business rules, and can also retrieve them by meaning (semantic search).
  • Faster responses: first word in 1.5 to 2 s instead of ~7 s on a new conversation.
  • More reliable: automatic correction of attribute casing, more robust queries, no more answers built on a failed result, and fixes derived from analysis of production traces.
  • Security: no query without a context filter, strengthened conversation guardrails, and confidentiality of sensitive data.
  • Agent configuration: new editors for instructions, starter questions and guardrails; instructions can be enabled or disabled; up to 12,000 characters; icon selection.
  • Interface: new Graphicient branding (animated logo) and more stable response streaming.