bill/whetstone_DSL
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add raw structural projector before candidate scoring (sprint 263)

Browse Source
This commit is contained in:
Bill
2026-02-26 16:59:28 -07:00
parent f534d7f0fa
commit fe85f1b515
8 changed files with 249 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

101
tools/mcp/project_raw_candidate_structure.py Normal file
View File

@@ -0,0 +1,101 @@
#!/usr/bin/env python3
import argparse
import json
from pathlib import Path
from typing import Dict, List
def load_json(path: Path):
with path.open("r", encoding="utf-8") as f:
return json.load(f)
def write_json(path: Path, obj) -> None:
with path.open("w", encoding="utf-8") as f:
json.dump(obj, f, indent=2, sort_keys=True)
f.write("\n")
def parse_gap_signals(backlog: Dict):
missing = [str(x.get("missing", "")) for x in (backlog.get("prioritized_missing") or [])]
ops = sorted({m.split(":", 1)[1] for m in missing if m.startswith("missing_prerequisite_op:")})
reasons = sorted({m.split(":", 1)[1] for m in missing if m.startswith("missing_reason_keyword:")})
contracts = sorted({m.split(":", 1)[1] for m in missing if m.startswith("missing_execution_contract:")})
min_count = 0
for m in missing:
if m.startswith("native_task_count<"):
try:
min_count = max(min_count, int(m.split("<", 1)[1]))
except ValueError:
pass
return ops, reasons, contracts, min_count
def normalize_task(task: Dict, ops: List[str], reasons: List[str], contracts: List[str]) -> Dict:
ec = dict(task.get("executionContract") or {})
for f in contracts:
ec[f] = True
ec.setdefault("deterministic", True)
ec.setdefault("rollbackRequired", True)
ec.setdefault("replayValidationRequired", True)
title = str(task.get("title", "Task")).strip() or "Task"
return {
**task,
"title": title,
"prerequisiteOps": sorted(set((task.get("prerequisiteOps") or []) + ops)),
"reasons": sorted(set((task.get("reasons") or []) + reasons + ["raw_structural_projection"])),
"executionContract": ec,
}
def expand_to_min_count(tasks: List[Dict], min_count: int, ops: List[str]) -> List[Dict]:
if min_count <= 0 or len(tasks) >= min_count or not tasks:
return tasks
out = list(tasks)
idx = 0
while len(out) < min_count:
base = dict(tasks[idx % len(tasks)])
tid = str(base.get("taskId", f"task-{idx+1}"))
op = ops[idx % len(ops)] if ops else "whetstone_validate_taskitem"
base["taskId"] = f"{tid}-p{len(out)+1}"
base["title"] = f"{base.get('title','Task')} [{op}]"
base["prerequisiteOps"] = sorted(set((base.get("prerequisiteOps") or []) + [op]))
out.append(base)
idx += 1
return out
def main() -> int:
p = argparse.ArgumentParser(description="Project raw candidate tasks into stronger structural shape before scoring.")
p.add_argument("--tasks", required=True)
p.add_argument("--top-gaps", required=True)
p.add_argument("--out", required=True)
p.add_argument("--out-report", required=True)
args = p.parse_args()
tasks = load_json(Path(args.tasks))
if not isinstance(tasks, list):
raise SystemExit("--tasks must be JSON array")
backlog = load_json(Path(args.top_gaps))
ops, reasons, contracts, min_count = parse_gap_signals(backlog)
projected = [normalize_task(t, ops, reasons, contracts) for t in tasks]
projected = expand_to_min_count(projected, min_count, ops)
report = {
"status": "ok",
"input_task_count": len(tasks),
"output_task_count": len(projected),
"required_ops_count": len(ops),
"required_reason_count": len(reasons),
"required_contract_count": len(contracts),
"target_min_task_count": min_count,
}
write_json(Path(args.out), projected)
write_json(Path(args.out_report), report)
print(json.dumps({"status": "ok", "input_task_count": len(tasks), "output_task_count": len(projected), "out": args.out}, sort_keys=True))
return 0
if __name__ == "__main__":
raise SystemExit(main())

67
tools/mcp/run_sprint_taskitem_pipeline.sh
View File

@@ -55,6 +55,7 @@ NATIVE_RAW_CANDIDATE_REQUIRE_UPLIFT="${WSTONE_NATIVE_RAW_CANDIDATE_REQUIRE_UPLIF
NATIVE_RAW_HARDEN_TOP_GAPS="${WSTONE_NATIVE_RAW_HARDEN_TOP_GAPS:-0}"
NATIVE_RAW_SIGNAL_TARGETED_VARIANTS="${WSTONE_NATIVE_RAW_SIGNAL_TARGETED_VARIANTS:-0}"
NATIVE_RAW_SIGNAL_TARGETED_MAX_VARIANTS="${WSTONE_NATIVE_RAW_SIGNAL_TARGETED_MAX_VARIANTS:-6}"
NATIVE_RAW_STRUCTURAL_PROJECTOR="${WSTONE_NATIVE_RAW_STRUCTURAL_PROJECTOR:-0}"
NATIVE_RAW_TOP_GAP_WEIGHTED_SELECT="${WSTONE_NATIVE_RAW_TOP_GAP_WEIGHTED_SELECT:-0}"
NATIVE_RAW_TOP_GAP_BACKLOG_FILE="${WSTONE_NATIVE_RAW_TOP_GAP_BACKLOG_FILE:-}"
NATIVE_RAW_TOP_GAP_REQUIRE_UPLIFT="${WSTONE_NATIVE_RAW_TOP_GAP_REQUIRE_UPLIFT:-0}"
@@ -130,6 +131,7 @@ NATIVE_MULTISHOT_JSON='{}'
NATIVE_RAW_CANDIDATE_SEARCH_JSON='{}'
NATIVE_RAW_HARDENING_JSON='{}'
NATIVE_RAW_SIGNAL_TARGETED_VARIANTS_JSON='{}'
NATIVE_RAW_STRUCTURAL_PROJECTOR_JSON='{}'
NATIVE_RAW_TOP_GAP_REQUIREMENTS_JSON='{}'
NATIVE_RAW_INTRINSIC_PROMPT_PACK_JSON='{}'
NATIVE_RAW_TEMPLATE_CONTROL_PACK_JSON='{}'
@@ -460,10 +462,17 @@ fi
if [[ "$NATIVE_RAW_CANDIDATE_SEARCH" == "1" ]]; then
top_gap_score_args=()
use_top_gap_weighted_select=false
raw_structural_projector_enabled=false
raw_structural_projector_projected_variants=0
baseline_projected_input_count=0
baseline_projected_output_count=0
if [[ "$NATIVE_RAW_TOP_GAP_WEIGHTED_SELECT" == "1" && -n "$NATIVE_RAW_TOP_GAP_BACKLOG_FILE" && -f "$NATIVE_RAW_TOP_GAP_BACKLOG_FILE" ]]; then
top_gap_score_args=(--top-gaps "$NATIVE_RAW_TOP_GAP_BACKLOG_FILE")
use_top_gap_weighted_select=true
fi
if [[ "$NATIVE_RAW_STRUCTURAL_PROJECTOR" == "1" && -n "$NATIVE_RAW_TOP_GAP_BACKLOG_FILE" && -f "$NATIVE_RAW_TOP_GAP_BACKLOG_FILE" ]]; then
raw_structural_projector_enabled=true
fi
python3 "$ROOT_DIR/tools/mcp/synthesize_raw_candidate_requirement_variants.py" \
--spec "$INPUT_FILE" \
--profiles "$NATIVE_IMPACT_COVERAGE_PROFILES" \
@@ -490,16 +499,31 @@ if [[ "$NATIVE_RAW_CANDIDATE_SEARCH" == "1" ]]; then
fi
printf '%s\n' "$TASKS" > "$OUT_DIR/02ae_candidate_0_tasks.json"
baseline_eval_tasks_path="$OUT_DIR/02ae_candidate_0_tasks.json"
baseline_eval_tasks_json="$TASKS"
if [[ "$raw_structural_projector_enabled" == true ]]; then
python3 "$ROOT_DIR/tools/mcp/project_raw_candidate_structure.py" \
--tasks "$OUT_DIR/02ae_candidate_0_tasks.json" \
--top-gaps "$NATIVE_RAW_TOP_GAP_BACKLOG_FILE" \
--out "$OUT_DIR/02ae_candidate_0_projected_tasks.json" \
--out-report "$OUT_DIR/02ae_candidate_0_projected_report.json" >/dev/null
baseline_eval_tasks_path="$OUT_DIR/02ae_candidate_0_projected_tasks.json"
baseline_eval_tasks_json="$(cat "$OUT_DIR/02ae_candidate_0_projected_tasks.json")"
baseline_projected_input_count="$(jq '.input_task_count // 0' "$OUT_DIR/02ae_candidate_0_projected_report.json")"
baseline_projected_output_count="$(jq '.output_task_count // 0' "$OUT_DIR/02ae_candidate_0_projected_report.json")"
raw_structural_projector_projected_variants=$((raw_structural_projector_projected_variants + 1))
fi
python3 "$ROOT_DIR/tools/mcp/score_native_tasks_profile_coverage.py" \
--spec "$INPUT_FILE" \
--profiles "$NATIVE_IMPACT_COVERAGE_PROFILES" \
--tasks "$OUT_DIR/02ae_candidate_0_tasks.json" \
--tasks "$baseline_eval_tasks_path" \
"${top_gap_score_args[@]}" \
--out "$OUT_DIR/02ae_candidate_0_score.json" >/dev/null
best_variant="0"
best_fail="$(jq '.failing_profile_count // 999' "$OUT_DIR/02ae_candidate_0_score.json")"
best_task_count="$(jq '.task_count // 0' "$OUT_DIR/02ae_candidate_0_score.json")"
best_top_gap_score="$(jq '.top_gap_score // 0' "$OUT_DIR/02ae_candidate_0_score.json")"
TASKS="$baseline_eval_tasks_json"
baseline_fail="$best_fail"
baseline_task_count="$best_task_count"
baseline_top_gap_score="$best_top_gap_score"
@@ -523,10 +547,22 @@ if [[ "$NATIVE_RAW_CANDIDATE_SEARCH" == "1" ]]; then
if [[ "$(printf '%s' "$cand_json" | jq -r '.success // false')" == "true" ]]; then
cand_tasks="$(printf '%s' "$cand_json" | jq '.tasks // []')"
printf '%s\n' "$cand_tasks" > "$OUT_DIR/02ae_candidate_${variant}_tasks.json"
cand_eval_tasks_path="$OUT_DIR/02ae_candidate_${variant}_tasks.json"
cand_eval_tasks_json="$cand_tasks"
if [[ "$raw_structural_projector_enabled" == true ]]; then
python3 "$ROOT_DIR/tools/mcp/project_raw_candidate_structure.py" \
--tasks "$OUT_DIR/02ae_candidate_${variant}_tasks.json" \
--top-gaps "$NATIVE_RAW_TOP_GAP_BACKLOG_FILE" \
--out "$OUT_DIR/02ae_candidate_${variant}_projected_tasks.json" \
--out-report "$OUT_DIR/02ae_candidate_${variant}_projected_report.json" >/dev/null
cand_eval_tasks_path="$OUT_DIR/02ae_candidate_${variant}_projected_tasks.json"
cand_eval_tasks_json="$(cat "$OUT_DIR/02ae_candidate_${variant}_projected_tasks.json")"
raw_structural_projector_projected_variants=$((raw_structural_projector_projected_variants + 1))
fi
python3 "$ROOT_DIR/tools/mcp/score_native_tasks_profile_coverage.py" \
--spec "$INPUT_FILE" \
--profiles "$NATIVE_IMPACT_COVERAGE_PROFILES" \
--tasks "$OUT_DIR/02ae_candidate_${variant}_tasks.json" \
--tasks "$cand_eval_tasks_path" \
"${top_gap_score_args[@]}" \
--out "$OUT_DIR/02ae_candidate_${variant}_score.json" >/dev/null
cand_fail="$(jq '.failing_profile_count // 999' "$OUT_DIR/02ae_candidate_${variant}_score.json")"
@@ -549,7 +585,7 @@ if [[ "$NATIVE_RAW_CANDIDATE_SEARCH" == "1" ]]; then
best_fail="$cand_fail"
best_task_count="$cand_task_count"
best_top_gap_score="$cand_top_gap_score"
TASKS="$cand_tasks"
TASKS="$cand_eval_tasks_json"
fi
successful=$((successful + 1))
fi
@@ -580,10 +616,22 @@ if [[ "$NATIVE_RAW_CANDIDATE_SEARCH" == "1" ]]; then
retry_success=true
retry_tasks="$(printf '%s' "$retry_json" | jq '.tasks // []')"
printf '%s\n' "$retry_tasks" > "$OUT_DIR/02af_raw_adaptive_retry_tasks.json"
retry_eval_tasks_path="$OUT_DIR/02af_raw_adaptive_retry_tasks.json"
retry_eval_tasks_json="$retry_tasks"
if [[ "$raw_structural_projector_enabled" == true ]]; then
python3 "$ROOT_DIR/tools/mcp/project_raw_candidate_structure.py" \
--tasks "$OUT_DIR/02af_raw_adaptive_retry_tasks.json" \
--top-gaps "$NATIVE_RAW_TOP_GAP_BACKLOG_FILE" \
--out "$OUT_DIR/02af_raw_adaptive_retry_projected_tasks.json" \
--out-report "$OUT_DIR/02af_raw_adaptive_retry_projected_report.json" >/dev/null
retry_eval_tasks_path="$OUT_DIR/02af_raw_adaptive_retry_projected_tasks.json"
retry_eval_tasks_json="$(cat "$OUT_DIR/02af_raw_adaptive_retry_projected_tasks.json")"
raw_structural_projector_projected_variants=$((raw_structural_projector_projected_variants + 1))
fi
python3 "$ROOT_DIR/tools/mcp/score_native_tasks_profile_coverage.py" \
--spec "$INPUT_FILE" \
--profiles "$NATIVE_IMPACT_COVERAGE_PROFILES" \
--tasks "$OUT_DIR/02af_raw_adaptive_retry_tasks.json" \
--tasks "$retry_eval_tasks_path" \
"${top_gap_score_args[@]}" \
--out "$OUT_DIR/02af_raw_adaptive_retry_score.json" >/dev/null
retry_fail="$(jq '.failing_profile_count // 999' "$OUT_DIR/02af_raw_adaptive_retry_score.json")"
@@ -594,7 +642,7 @@ if [[ "$NATIVE_RAW_CANDIDATE_SEARCH" == "1" ]]; then
best_fail="$retry_fail"
best_task_count="$retry_task_count"
best_top_gap_score="$retry_top_gap_score"
TASKS="$retry_tasks"
TASKS="$retry_eval_tasks_json"
retry_applied=true
fi
fi
@@ -611,6 +659,12 @@ if [[ "$NATIVE_RAW_CANDIDATE_SEARCH" == "1" ]]; then
else
NATIVE_RAW_ADAPTIVE_RETRY_JSON='{"enabled":false}'
fi
NATIVE_RAW_STRUCTURAL_PROJECTOR_JSON="$(jq -nc \
--argjson enabled "$raw_structural_projector_enabled" \
--argjson projected_variant_count "$raw_structural_projector_projected_variants" \
--argjson baseline_input_task_count "$baseline_projected_input_count" \
--argjson baseline_output_task_count "$baseline_projected_output_count" \
'{enabled:$enabled, projected_variant_count:$projected_variant_count, baseline_input_task_count:$baseline_input_task_count, baseline_output_task_count:$baseline_output_task_count}')"
NATIVE_RAW_CANDIDATE_SEARCH_JSON="$(jq -nc \
--argjson enabled true \
@@ -649,6 +703,7 @@ if [[ "$NATIVE_RAW_CANDIDATE_SEARCH" == "1" ]]; then
else
NATIVE_RAW_CANDIDATE_SEARCH_JSON='{"enabled":false}'
NATIVE_RAW_SIGNAL_TARGETED_VARIANTS_JSON='{"enabled":false}'
NATIVE_RAW_STRUCTURAL_PROJECTOR_JSON='{"enabled":false}'
fi
if [[ "$NATIVE_RAW_HARDEN_TOP_GAPS" == "1" ]]; then
printf '%s\n' "$TASKS" > "$OUT_DIR/02af_raw_hardening_input_tasks.json"
@@ -997,6 +1052,7 @@ SUMMARY_JSON="$(jq -nc \
--argjson native_raw_intrinsic_prompt_pack "$NATIVE_RAW_INTRINSIC_PROMPT_PACK_JSON" \
--argjson native_raw_template_control_pack "$NATIVE_RAW_TEMPLATE_CONTROL_PACK_JSON" \
--argjson native_raw_signal_targeted_variants "$NATIVE_RAW_SIGNAL_TARGETED_VARIANTS_JSON" \
--argjson native_raw_structural_projector "$NATIVE_RAW_STRUCTURAL_PROJECTOR_JSON" \
--argjson native_raw_candidate_search "$NATIVE_RAW_CANDIDATE_SEARCH_JSON" \
--argjson native_raw_adaptive_retry "$NATIVE_RAW_ADAPTIVE_RETRY_JSON" \
--argjson native_raw_hardening "$NATIVE_RAW_HARDENING_JSON" \
@@ -1031,6 +1087,7 @@ SUMMARY_JSON="$(jq -nc \
native_raw_intrinsic_prompt_pack: $native_raw_intrinsic_prompt_pack,
native_raw_template_control_pack: $native_raw_template_control_pack,
native_raw_signal_targeted_variants: $native_raw_signal_targeted_variants,
native_raw_structural_projector: $native_raw_structural_projector,
native_raw_candidate_search: $native_raw_candidate_search,
native_raw_adaptive_retry: $native_raw_adaptive_retry,
native_raw_hardening: $native_raw_hardening,