Handbook

Forge Intelligence owns problem-class reasoning economy

Accepted for v0 scaffold.

Updated

Status

Accepted for v0 scaffold.

Context

LCDL and lmeta provide deterministic-first orchestration with opt-in LLM break-in. Consumers implement local routing ladders independently. Token spend and ambiguity are documented in Platform standouts but not standardized as attach contracts on the ForgeRun spine.

Platform documentation also uses "intelligence layer" for hydration claim registries — a separate concern from runtime reasoning economy.

Decision

Add forge-intelligence as a private ecosystem product (abbr fi) with handbook shell (fiw). It owns problem-class packs, Human-Life bundle import, graph orchestration, concept + session dual-wiki, and attach schemas:

v1 (session attach):

  • intelligence_policy.v1
  • ambiguity_assessment.v1
  • token_ledger.v1
  • route_trace.v1

v2 (manifest + graph, in forge-intelligence/schemas/v2/):

  • intelligence_manifest.ref
  • task_graph
  • maturity_axes
  • artifact_ref

Schemas live in forge-intelligence/schemas/ for v0. Platform boundaries and ADRs reference them; promotion into forge-platform/schemas/ follows demonstrated consumer integration.

Forge Intelligence composes LCDL/lmeta — it does not fork the executor. Intelligence-specific LCDL contracts and lmeta units live under forge-intelligence/lcdl/ and forge-intelligence/lmeta/, not in forge-lcdl.

Consequences

  • Reasoning economy is a named product with clear boundaries in product-boundaries.md.
  • LCDL remains the governed reasoning layer (ADR-0004).
  • Doc hydration docs must qualify "intelligence" when cross-linking.
  • Lenses remains the system of record for runs and human review.

References

  • forge-intelligence/docs/adr/ADR-0001-intelligence-as-reasoning-economy.md
  • forge-platform/product-repo-stubs/intelligence/docs/forge/product-boundary.md