Skip to content

Latera

Target

There is no executable end-to-end Latera pipeline in this repository.

Latera is a target lateral-reasoning capability. No executable end-to-end Latera pipeline, operator registry, transfer classifier or editable thesis graph is implemented in this repository. Ordinary legal search and the generic research planner do not establish that capability.

flowchart LR
  Case["Case and seed authorities"] --> Principles["Candidate transferable principles"]
  Principles --> Search["Cross-domain legal discovery"]
  Search --> Read["Read original reasoning"]
  Read --> Assess["Assess transfer and limitations"]
  Assess --> Output["Reviewable theses and supporting evidence"]

Each arrow above is a target step. Cross-domain candidate discovery depends on specialist Legal Core work; task orchestration and transfer assessment would belong to product-worker.

A semantically similar judgment can arise under incompatible legal conditions. The target assessment therefore needs the original reasoning, the proposed bridge to the current case, and explicit limits. A similarity score alone must not become a legal success probability.

A risk classification would need deterministic validation of its evidence and assessment inputs. A prompt asking for a “green” classification does not enforce those conditions. That validator and its tests remain unimplemented.

Reference candidate budgets, shortlist thresholds and ordering weights belong in a configuration or golden-fixture file consumed by the future implementation and tests. There is no current Latera configuration to link to; this page deliberately makes no claim of parity with the legacy pipeline.

For capabilities available today, see Legal research and Agent execution.