Implement Quick Evaluation and Quality Scoring for Path QA

- Added `_quick_evaluate_steps_qa` function to streamline path quality assessment without recursive API calls, enhancing performance for slot comparisons.
- Introduced `compute_deterministic_path_quality_score` to provide a heuristic quality score based on gaps and off-topic steps, improving evaluation accuracy.
- Updated `_run_unified_slot_improvement_review` to utilize the new quick evaluation method, optimizing the review process and integrating quality scoring.
- Enhanced `build_path_qa_summary` to include quality score calculations, ensuring comprehensive feedback on path evaluations.
- Refactored related functions for improved clarity and efficiency in handling path quality assessments.

backend/planning_exercise_path_builder.py

@@ -36,6 +36,7 @@ from planning_stage_context import build_contextualized_stage_goal, resolve_path
 from planning_exercise_path_qa import (
     apply_llm_path_reorder,
     build_path_qa_summary,
+    compute_deterministic_path_quality_score,
     detect_off_topic_steps,
     detect_path_gaps,
     insert_bridge_exercises,
@@ -2399,6 +2400,110 @@ def _evaluate_steps_for_compare_qa(
     return suggest_progression_path(cur, tenant=tenant, body=eval_body)
 
 
+def _quick_evaluate_steps_qa(
+    cur,
+    *,
+    goal_query: str,
+    semantic_brief: PlanningSemanticBrief,
+    steps: Sequence[Mapping[str, Any]],
+    roadmap_ctx: Optional[ProgressionRoadmapContext],
+) -> Dict[str, Any]:
+    """Schnelle Pfad-QS ohne rekursiven API-Lauf — für Slot-Vergleiche."""
+    roadmap_first = roadmap_ctx is not None
+    steps_list = list(steps or [])
+    gaps = detect_path_gaps(
+        cur,
+        steps_list,
+        brief=semantic_brief,
+        roadmap_first=roadmap_first,
+    )
+    off_topic_steps = detect_off_topic_steps(
+        cur,
+        steps_list,
+        brief=semantic_brief,
+        goal_query=goal_query,
+    )
+    multistage_qa = run_multistage_path_qa(
+        off_topic_steps=off_topic_steps,
+        stripped_off_topic=[],
+        gaps=gaps,
+        llm_qa=None,
+        llm_applied=False,
+    )
+    path_qa = build_path_qa_summary(
+        gaps=gaps,
+        bridge_inserts=[],
+        ai_proposals=[],
+        gap_fill_offers=[],
+        off_topic_steps=off_topic_steps,
+        stripped_off_topic=[],
+        llm_qa=None,
+        llm_applied=False,
+        roadmap_qa_mode="roadmap_first_lite" if roadmap_first else None,
+        multistage_qa=multistage_qa,
+    )
+    if path_qa.get("quality_score") is None:
+        path_qa["quality_score"] = compute_deterministic_path_quality_score(
+            gaps=gaps,
+            off_topic_steps=off_topic_steps,
+            steps=steps_list,
+            multistage_qa=multistage_qa,
+        )
+    return path_qa
+
+
+def _off_topic_slot_indices(path_qa: Optional[Mapping[str, Any]]) -> Set[int]:
+    return set(_off_topic_reasons_by_slot((path_qa or {}).get("off_topic_steps") or []).keys())
+
+
+def _slot_suggestion_accepted(
+    *,
+    baseline_qa: Optional[Mapping[str, Any]],
+    projected_qa: Optional[Mapping[str, Any]],
+    baseline_score: Optional[float],
+    projected_score: Optional[float],
+    diff: Mapping[str, Any],
+    off_topic: bool,
+    major_idx: int,
+) -> bool:
+    """Entscheidet, ob ein Slot-Vorschlag in die Liste kommt."""
+    base_id = diff.get("baseline_exercise_id")
+    prop_id = diff.get("proposed_exercise_id")
+    base_off = _off_topic_slot_indices(baseline_qa)
+    proj_off = _off_topic_slot_indices(projected_qa)
+
+    if off_topic and base_id is not None:
+        if major_idx in base_off and major_idx not in proj_off:
+            return True
+        if major_idx in base_off and prop_id is not None:
+            return _slot_diff_improves_path(diff, _quality_delta(baseline_score, projected_score), off_topic=True)
+
+    if base_id is None and prop_id is not None:
+        return _slot_diff_improves_path(diff, _quality_delta(baseline_score, projected_score), off_topic=False)
+
+    if base_id is not None and prop_id is not None:
+        if int(base_id) == int(prop_id):
+            return False
+        return _slot_diff_improves_path(diff, _quality_delta(baseline_score, projected_score), off_topic=False)
+
+    if base_id is None and prop_id is None and diff.get("proposed_is_ai_proposal"):
+        return _slot_diff_improves_path(
+            diff,
+            _quality_delta(baseline_score, projected_score),
+            off_topic=off_topic or major_idx in base_off,
+        )
+    return False
+
+
+def _quality_delta(
+    baseline_score: Optional[float],
+    projected_score: Optional[float],
+) -> Optional[float]:
+    if baseline_score is None or projected_score is None:
+        return None
+    return round(float(projected_score) - float(baseline_score), 4)
+
+
 def _apply_slot_diff_to_steps(
     baseline_steps: Sequence[Mapping[str, Any]],
     diff: Mapping[str, Any],
@@ -2784,6 +2889,14 @@ def _run_unified_slot_improvement_review(
     )
     baseline_steps = list(qa_pack.get("steps") or baseline_steps)
     baseline_qa = qa_pack.get("path_qa") if isinstance(qa_pack.get("path_qa"), dict) else {}
+    if baseline_qa.get("quality_score") is None:
+        baseline_qa = dict(baseline_qa)
+        baseline_qa["quality_score"] = compute_deterministic_path_quality_score(
+            gaps=baseline_qa.get("large_gaps") or [],
+            off_topic_steps=baseline_qa.get("off_topic_steps") or [],
+            steps=baseline_steps,
+            multistage_qa=baseline_qa,
+        )
     baseline_score = _path_qa_quality_score(baseline_qa)
     gap_fill_offers = list(qa_pack.get("gap_fill_offers") or [])
     off_topic_map = _off_topic_reasons_by_slot(baseline_qa.get("off_topic_steps") or [])
@@ -2794,14 +2907,6 @@ def _run_unified_slot_improvement_review(
 
     suggestions: List[Dict[str, Any]] = []
     rejected: List[Dict[str, Any]] = []
-    scored_eval_body = body.model_copy(
-        update={
-            "include_llm_path_qa": False,
-            "include_ai_gap_fill": False,
-            "auto_rematch_after_qa": False,
-            "include_roadmap_preview": False,
-        }
-    )
 
     for step_index, stage_spec in enumerate(roadmap_ctx.stage_specs):
         major_idx = int(stage_spec.major_step_index)
@@ -2856,6 +2961,7 @@ def _run_unified_slot_improvement_review(
             anchor_variant_id=anchor_variant_id,
             used=used_other,
             exclude_exercise_id=exclude_id if not off_topic else int(current_id) if current_id else None,
+            max_candidates=3,
         )
 
         accepted_for_slot = False
@@ -2882,22 +2988,25 @@ def _run_unified_slot_improvement_review(
                 if int(raw.get("roadmap_major_step_index", -1)) == major_idx:
                     merged_steps[i] = {**raw, **candidate, "roadmap_major_step_index": major_idx}
                     break
-            eval_res = _evaluate_steps_for_compare_qa(
+            eval_res = _quick_evaluate_steps_qa(
                 cur,
-                tenant=tenant,
+                goal_query=goal_query,
-                body=scored_eval_body,
+                semantic_brief=semantic_brief,
                 steps=merged_steps,
+                roadmap_ctx=roadmap_ctx,
             )
-            projected_qa = (
+            projected_qa = eval_res if isinstance(eval_res, dict) else None
-                eval_res.get("path_qa")
-                if isinstance(eval_res, dict) and isinstance(eval_res.get("path_qa"), dict)
-                else None
-            )
             projected_score = _path_qa_quality_score(projected_qa)
-            delta: Optional[float] = None
+            delta = _quality_delta(baseline_score, projected_score)
-            if baseline_score is not None and projected_score is not None:
+            improves = _slot_suggestion_accepted(
-                delta = round(projected_score - baseline_score, 4)
+                baseline_qa=baseline_qa,
-            improves = _slot_diff_improves_path(diff_stub, delta, off_topic=off_topic)
+                projected_qa=projected_qa,
+                baseline_score=baseline_score,
+                projected_score=projected_score,
+                diff=diff_stub,
+                off_topic=off_topic,
+                major_idx=major_idx,
+            )
             suggestion_type = (
                 "remove_and_replace"
                 if off_topic and current_id is not None
@@ -2990,24 +3099,25 @@ def _run_unified_slot_improvement_review(
             "proposed_title": ai_step.get("title"),
         }
         merged_steps = _apply_slot_diff_to_steps(baseline_steps, diff_stub, [ai_step])
-        eval_res = _evaluate_steps_for_compare_qa(
+        eval_res = _quick_evaluate_steps_qa(
             cur,
-            tenant=tenant,
+            goal_query=goal_query,
-            body=scored_eval_body,
+            semantic_brief=semantic_brief,
             steps=merged_steps,
+            roadmap_ctx=roadmap_ctx,
         )
-        projected_qa = (
+        projected_qa = eval_res if isinstance(eval_res, dict) else None
-            eval_res.get("path_qa")
-            if isinstance(eval_res, dict) and isinstance(eval_res.get("path_qa"), dict)
-            else None
-        )
         projected_score = _path_qa_quality_score(projected_qa)
-        delta = (
+        delta = _quality_delta(baseline_score, projected_score)
-            round(projected_score - baseline_score, 4)
+        improves = _slot_suggestion_accepted(
-            if baseline_score is not None and projected_score is not None
+            baseline_qa=baseline_qa,
-            else None
+            projected_qa=projected_qa,
+            baseline_score=baseline_score,
+            projected_score=projected_score,
+            diff=diff_stub,
+            off_topic=off_topic or major_idx in _off_topic_slot_indices(baseline_qa),
+            major_idx=major_idx,
         )
-        improves = _slot_diff_improves_path(diff_stub, delta, off_topic=off_topic or current_id is None)
         entry = {
             **diff_stub,
             "baseline_slot_status": current.get("slot_status"),

backend/planning_exercise_path_qa.py

@@ -745,12 +745,44 @@ def build_path_qa_summary(
                 f"Schritt „{o.get('title')}“ passt nicht zum Pfad-Thema"
                 for o in off_topic
             ]
+        summary["quality_score"] = compute_deterministic_path_quality_score(
+            gaps=gaps,
+            off_topic_steps=off_topic,
+            steps=steps,
+            multistage_qa=multistage_qa,
+        )
     return summary
 
 
+def compute_deterministic_path_quality_score(
+    *,
+    gaps: Sequence[Mapping[str, Any]],
+    off_topic_steps: Sequence[Mapping[str, Any]],
+    steps: Optional[Sequence[Mapping[str, Any]]] = None,
+    multistage_qa: Optional[Mapping[str, Any]] = None,
+) -> float:
+    """Heuristische Pfad-QS ohne LLM — Basis für Slot-Vergleiche."""
+    score = 0.92
+    score -= 0.08 * len(off_topic_steps or [])
+    score -= 0.05 * len(gaps or [])
+    if steps:
+        empty = sum(
+            1
+            for s in steps
+            if isinstance(s, dict)
+            and s.get("exercise_id") is None
+            and not s.get("is_ai_proposal")
+        )
+        score -= 0.06 * empty
+    hint_count = int((multistage_qa or {}).get("optimization_hint_count") or 0)
+    score -= min(0.14, 0.02 * hint_count)
+    return max(0.35, min(0.98, round(score, 4)))
+
+
 __all__ = [
     "apply_llm_path_reorder",
     "build_path_qa_summary",
+    "compute_deterministic_path_quality_score",
     "detect_off_topic_steps",
     "detect_path_gaps",
     "is_roadmap_planned_neighbor_pair",

backend/tests/test_planning_deterministic_quality_score.py

@@ -0,0 +1,21 @@
+"""Deterministische Pfad-QS ohne LLM."""
+from planning_exercise_path_qa import compute_deterministic_path_quality_score
+
+
+def test_deterministic_quality_score_penalizes_off_topic():
+    base = compute_deterministic_path_quality_score(gaps=[], off_topic_steps=[])
+    with_off = compute_deterministic_path_quality_score(
+        gaps=[],
+        off_topic_steps=[{"roadmap_major_step_index": 1}],
+    )
+    assert with_off < base
+
+
+def test_deterministic_quality_score_penalizes_empty_slots():
+    base = compute_deterministic_path_quality_score(gaps=[], off_topic_steps=[], steps=[])
+    with_empty = compute_deterministic_path_quality_score(
+        gaps=[],
+        off_topic_steps=[],
+        steps=[{"exercise_id": None}, {"exercise_id": 1}],
+    )
+    assert with_empty < base

frontend/src/components/ProgressionGraphEditor.jsx

@@ -500,10 +500,11 @@ export default function ProgressionGraphEditor({ graphId, embedded = false, onSa
       unified_slot_review: true,
       baseline_evaluate_steps: slotsToEvaluateSteps(synced),
       include_llm_intent: false,
-      include_llm_path_qa: false,
       auto_rematch_after_qa: false,
     })
-    setPathQa(reviewRes?.path_qa || null)
+    const qa = reviewRes?.path_qa || null
+    setPathQa(qa)
+    setDraft((prev) => (prev ? { ...prev, lastFindings: qa } : prev))
 
     const compareRes = buildProgressionComparePayload(null, reviewRes)
     setGapFillOffers(mergeGapOffersForDraft(synced, reviewRes, reviewRes))