feat: keep an unchanged description, list conditions, review coverage with Jev

After case 02 (completed at 100%, but with two Revisions of the Target
architecture for gaps validation found one pass at a time):

- A Deliverable stands while what it was written from is unchanged. The
  architecture description is written from the Elements and Relationships,
  so a Revision that only adds or rewords Checks no longer runs the describer.
- Non-functional requirements and recommendations list their conditions,
  one entry per separate thing that must be true; submission requires them.
- A coverage review at submission (JEV_COVERAGE_REVIEW, off by default):
  for each condition, Jev chooses which Check verifies it, or none. Asked as
  one broad yes-or-no per requirement it could not tell a complete design
  from one with known gaps; asked as a choice per condition it raised all
  four known gaps of case 02 and nothing on the complete graph. An opt-in
  live test (LIVE_JEV=1) repeats that.
- The validator is told to report every gap in its first pass.

Plans record the case 02 run and both forms of the review with figures.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-10-07 20:58:38 +01:00
parent f1f6f1b83a
commit 5adc911a82
29 changed files with 1687 additions and 16 deletions

View File

@@ -28,6 +28,7 @@ from app.adk.runtime import (
RoleSpec,
RunLimiter,
Stopped,
SubmissionReview,
Submitted,
ToolFactory,
ToolSpec,
@@ -61,6 +62,9 @@ from app.workflows.deliverables import (
logger = logging.getLogger(__name__)
# Working-memory key: the last submission of a Deliverable with what it was written from.
STANDING = "standing"
NO_USER_REPLY = (
"No user is available. State your assumption, record it in the Deliverable under "
"'Assumptions', and proceed."
@@ -98,6 +102,7 @@ def make_execution_nodes(
limiter: RunLimiter | None = None,
skill_versions: Mapping[str, str] | None = None,
orchestrator: OrchestratorPolicy | None = None,
submission_review: SubmissionReview | None = None,
) -> ExecutionNodes:
tool_factory = tool_factory or _no_tools
# How an Orchestrator turn is decided. The default drives the tools from the
@@ -357,6 +362,48 @@ def make_execution_nodes(
working[workspace.FILES] = dict(
upstream.get(source.value, {}).get("attachments", {})
)
spec = DEPENDENCY_GRAPH[kind]
# Feedback on this Deliverable itself. A note that a dependency was revised
# is not a reason to write it again; the basis says whether that matters.
feedback = [
item
for item in task_input["context"].get("feedback") or []
if not str(item).startswith("Dependency ")
]
basis = spec.written_from(upstream) if spec.written_from else None
standing = working.get(STANDING)
if (
basis
and isinstance(standing, Mapping)
and standing.get("basis") == basis
and standing.get("feedback") == feedback
):
# What it was written from has not changed, so the last submission
# still stands: no Specialist is run.
logger.info(
"[%s] %s kept %s (attempt %d): written from an unchanged basis",
task_input["execution_id"],
role.name,
kind,
task_input["attempt"],
)
kept = Submitted(standing["content"], dict(standing["attachments"]))
return {
"outcomes": [
{
**_outcome_record(role.name, kind, task_input["attempt"], kept),
"refusals": [],
}
],
"working": {kind.value: working},
}
async def review(attachments: Mapping[str, str]) -> Sequence[str]:
assert submission_review is not None
return await submission_review(
kind, attachments, upstream, task_input["execution_id"]
)
refusals: list[dict[str, Any]] = []
tools = _noting_refusals(
specialist_tools(
@@ -366,6 +413,7 @@ def make_execution_nodes(
upstream=upstream,
working=working,
library=task_input.get("library"),
review=review if submission_review else None,
),
refusals,
role.name,
@@ -402,6 +450,13 @@ def make_execution_nodes(
time.monotonic() - started,
_summarise_outcome(outcome),
)
if basis and isinstance(outcome, Submitted):
working[STANDING] = {
"basis": basis,
"feedback": feedback,
"content": outcome.content,
"attachments": dict(outcome.attachments),
}
return {
"outcomes": [
{