Files

T

CI / build (push) Has been cancelled

Details

Flatten model-divergence subtasks into 3 independent tasks

Replaced parent+subtask structure with 3 standalone tasks:
- mde-manifest-detection (foundational: models.json schema, detect_models.py)
- mde-interactive-enforcement (conflict matrix, --model args, audit, badges)
- mde-loop-enforcement (loop.json per-role model, {model} substitution, check-gate)

Parent archived to tasks/complete/ for history. Each task has its own
SPEC.md and is at research:awaiting_approval.

Rationale: subtasks are independently actionable with their own lifecycle.
Parent container added complexity without benefit.

2026-06-25 07:25:10 -04:00

2.3 KiB

Raw Blame History

SPEC — mde-loop-enforcement

Problem

Loops have no model-divergence enforcement. A loop could use the same model for both implementation and verification, defeating the purpose of the loop verifier gate. Loop runners need per-role model binding and {model} substitution in harness commands.

Design

Full design in design/framework/functional.md §4 and design/framework/technical.md §§10, §13.

Key decisions

Per-role model: loop.json roles section with optional model field per role. Defaults to models.json default model if not specified.
Substitution: {model} placeholder in loop.json harness.command gets replaced with the assigned model for that role.
Check-gate: New model-divergence brake gate: loop-verify model must differ from loop-implement model in multi-LLM mode. Single-LLM mode: advisory only.

Scope

scripts/loop-runner.py:
- _invoke_harness (line 366-404): {model} placeholder substitution from loop.json role config
scripts/status.py:
- --check-gate: model-divergence brake gate (loop-verify ≠ loop-implement in multi-LLM mode)
- cmd_create_loop / cmd_install_schedule: validate loop.json per-role model fields
templates/loops/: update loop templates with roles schema example
design/loops/technical.md: document {model} substitution
tests/test_model_divergence.py: loop model binding, {model} substitution, check-gate halt

Dependencies

Depends on mde-interactive-enforcement (needs CONFLICT_MATRIX and _check_conflict)

Out of scope

Interactive --transition --model / --claim --model (already done in mde-interactive-enforcement)
Rule agents (FW-2, FW-3)
Agent tab redesign (FW-1)

Success criteria

loop.json with roles.implementer.model → _invoke_harness substitutes {model} in harness command
loop.json without per-role model → defaults to models.json default model
--check-gate in multi-LLM mode halts if loop-verify model = loop-implement model
--check-gate in single-LLM mode does NOT halt (advisory only)
pytest tests/ -q green

References

design/framework/functional.md §4 (Model-Divergence Enforcement)
design/framework/technical.md §§10 (harness invocation), §13 (loop integration)
design/loops/technical.md (loop runner architecture)

2.3 KiB Raw Blame History