Skip to content

v1.0 / v2.0 Capability Matrix

Version position: v2.0

Purpose: explain how v2.0 follows v1.0 while retaining the foundational semantic-layer contract.

v2.0 is not a deprecation statement for v1.0. TM/QM, JSON Query DSL, MCP tools, governance, and query evidence remain the foundation. v2.0 adds Java-first complex analysis, trace, Memory Grid, Pivot, and Experience Recipe capabilities.

Overall relationship

Dimensionv1.0v2.0Upgrade guidance
Documentation roleFoundational semantic-layer whitepaperAdditional-capability whitepaperKeep v1.0 as the syntax/concept entry point
Implementation positionImplemented foundational capabilityJava-first additionsDo not infer parity for every language or deployment
BoundarySemantic query, MCP, Compose, governance, evidenceComplex analysis, trace, Memory Grid, Pivot, RecipeAdopt each addition as a signed contract
Release relationshipFrozen maintenanceFollow-on evolutionDo not append new capabilities to v1.0

Capability comparison

Areav1.0v2.0Compatibility
TM/QM modelingCoveredInheritedv2.0 does not rewrite the foundational syntax
JSON Query DSLCoveredInherited as the base query surfaceExisting request models do not need rewriting
MCP toolsDiscovery, description, query, compositionJava MCP redispatch and trace evidenceExisting entry points remain; debug/evidence fields may be added
GovernanceField validation, model visibility, row slicesExtended to Recipe, Memory Grid, and Pivot contractsv2.0 does not relax policy boundaries
Query evidencePayload, SQL, status, result sampleTrace, stage plan, contract, and Recipe evidenceIntegrations should tolerate additional debug information
ComposeMulti-step query and intermediate-result analysisCooperates with DSL_CTE and Memory Grid layersPrefer signed stage plans for complex analysis
DSL_CTENot a v1.0 primary capabilityNarrow staged-analysis contractUse signed stages and bridge templates only
Governed ExpressionBasic calculated fields and query-time calculationsSigned result-stage formula subsetDo not replace it with a free expression language
Period-over-periodFollow-up or local capabilityNarrow month/quarter comparison templatesDo not generalize to arbitrary calendars
Funnel/attributionFollow-up or local capabilityCRM funnel, target-event window, target-month bucket templatesNot a generic funnel
Memory GridComposition directionBounded secondary analysis, handle lifecycle, alignment contractBounded result handles only
PivotBasic or component-level capabilityTree axis, axis window, cascade drilldown contractNot a complete BI pivot product
Experience RecipeFollow-up directionExact registry lookup, governance filter, publish gate, signature SPINot a marketplace or vector search
Database dialectsAbstracted dialect foundationAdditional complex-analysis parity evidenceNarrow evidence is not a full release gate
Python parityMay have an independent implementationv2.0 is Java-firstCheck parity records separately

Integration impact

Existing v1.0 integrations can keep:

  • TM/QM model structures;
  • basic JSON Query DSL;
  • basic MCP list/describe/query calls;
  • basic policy and row-slice strategies.

Recommended additions:

  • display or record trace IDs, debug evidence, and contract evidence;
  • introduce signed DSL_CTE, Memory Grid, or Pivot contracts for recurring complex analysis instead of expanding free prompts;
  • use Experience Recipe for recurring analysis while retaining namespace, tenant, permission, and owner filtering;
  • clarify or reject unsigned scenarios rather than falling back to bare SQL.

Migration order

  1. Enable non-blocking trace and evidence display.
  2. Validate signed DSL_CTE stages in read-only scenarios.
  3. Pilot signed result-stage formulas in a small set of use cases.
  4. Enable contract evidence for explicitly supported Pivot shapes.
  5. Introduce Memory Grid alignment contracts for bounded cross-model results.
  6. Capture recurring complex analysis as Experience Recipes.

Incompatibility risks

  • Treating Java-first v2.0 behavior as synchronized across Python or every deployment.
  • Treating a signed contract as a generalized capability.
  • Ignoring unsigned shapes in debug evidence.
  • Allowing a Recipe to bypass model, policy, or validator checks.
  • Treating Memory Grid as an unbounded in-memory SQL engine.
  • Treating a Pivot contract as a complete frontend BI product.

Conclusion

v1.0 remains the foundational semantic-layer contract; v2.0 describes additional complex-analysis capability. Upgrade by inheriting the foundation, signing each addition, making evidence visible, and failing closed.