Phase N+O: 컨테이너 기반 레이아웃 + Step B 제거 + 전면 정리

- Phase N: catalog 개선, fallback 전면 제거, Kei API 무한 재시도, topic_id 버그 수정 - Phase O: 컨테이너 스펙 계산(비중→px), 블록 스펙 확정, 렌더러 container div - Step B(Sonnet) 제거: Kei(A-2)+코드로 대체. STEP_B_PROMPT/fallback/DOWNGRADE_MAP 삭제 - Selenium: container div 감지 추가 - catalog.yaml: ref_chars 구조 변환 + FAISS 재빌드 - 문서 전면 갱신: README, PROGRESS, IMPROVEMENT, Phase I~O md Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-27 15:20:51 +09:00
parent ffad1ba82a
commit b0bcffc0f6
28 changed files with 8450 additions and 1530 deletions
--- a/src/space_allocator.py
+++ b/src/space_allocator.py
@@ -0,0 +1,312 @@
+"""Phase O: 컨테이너 기반 공간 할당 시스템.
+
+Kei 비중 → 컨테이너 px 확정 → 블록 제약 계산 → 편집자 스펙 생성.
+LLM 추정이 아닌 결정론적 계산.
+
+주요 함수:
+- calculate_container_specs(): Kei 비중 → 역할별 ContainerSpec
+- finalize_block_specs(): 컨테이너 크기 → 블록별 내부 스펙
+- calculate_trim_chars(): 초과 px → 삭제 글자 수
+"""
+from __future__ import annotations
+
+import math
+import logging
+from dataclasses import dataclass, field
+from typing import Any
+
+logger = logging.getLogger(__name__)
+
+
+# ──────────────────────────────────────
+# height_cost → px 범위 매핑
+# ──────────────────────────────────────
+HEIGHT_COST_PX_RANGE = {
+    "compact": (30, 80),
+    "medium": (80, 200),
+    "large": (200, 350),
+    "xlarge": (350, 500),
+}
+
+HEIGHT_COST_ORDER = {"compact": 0, "medium": 1, "large": 2, "xlarge": 3}
+
+# 역할별 zone 매핑 (기본)
+ROLE_ZONE_MAP = {
+    "본심": "body",
+    "배경": "body",
+    "첨부": "sidebar",
+    "결론": "footer",
+}
+
+# 폰트 설정 기본값
+DEFAULT_FONT_SIZE_PX = 15.2
+DEFAULT_LINE_HEIGHT = 1.7
+DEFAULT_AVG_CHAR_WIDTH_PX = 14.4  # fonttools 실측 기반 (Pretendard 한글)
+
+
+# ──────────────────────────────────────
+# ContainerSpec 데이터 클래스
+# ──────────────────────────────────────
+@dataclass
+class ContainerSpec:
+    """역할별 컨테이너 스펙."""
+    role: str               # "본심", "배경", "첨부", "결론"
+    zone: str               # "body", "sidebar", "footer"
+    topic_ids: list[int]    # 이 컨테이너에 속하는 topic ID들
+    weight: float           # Kei가 판단한 비중 (0.0~1.0)
+    height_px: int          # 컨테이너 높이 (px)
+    width_px: int           # 컨테이너 너비 (px)
+    max_height_cost: str    # 허용 최대 height_cost ("compact"/"medium"/"large"/"xlarge")
+    block_constraints: dict = field(default_factory=dict)  # 블록 내부 제약
+
+
+# ──────────────────────────────────────
+# O-1: 컨테이너 스펙 계산
+# ──────────────────────────────────────
+def calculate_container_specs(
+    page_structure: dict[str, Any],
+    topics: list[dict[str, Any]],
+    preset: dict[str, Any],
+    slide_width: int = 1280,
+    slide_height: int = 720,
+    gap_px: int = 20,
+) -> dict[str, ContainerSpec]:
+    """Kei 비중 → 역할별 ContainerSpec 계산.
+
+    결정론적. AI 호출 없음.
+
+    Args:
+        page_structure: Kei 판단 {"본심": {"topic_ids": [3], "weight": 0.6}, ...}
+        topics: 각 topic의 purpose, role, layer
+        preset: 프리셋 zone 정보 (budget_px, width_pct)
+        slide_width: 슬라이드 너비 (px)
+        slide_height: 슬라이드 높이 (px)
+        gap_px: 컨테이너 간 간격 (px)
+
+    Returns:
+        {"본심": ContainerSpec(...), "배경": ContainerSpec(...), ...}
+    """
+    zones = preset.get("zones", {})
+    specs: dict[str, ContainerSpec] = {}
+
+    # zone별로 해당 역할들의 비중 합산
+    zone_roles: dict[str, list[tuple[str, dict]]] = {}  # zone → [(role, info), ...]
+    for role_name, role_info in page_structure.items():
+        if not isinstance(role_info, dict):
+            continue
+        zone = ROLE_ZONE_MAP.get(role_name, "body")
+        if zone not in zone_roles:
+            zone_roles[zone] = []
+        zone_roles[zone].append((role_name, role_info))
+
+    for zone_name, role_list in zone_roles.items():
+        zone_info = zones.get(zone_name, {})
+        zone_budget = zone_info.get("budget_px", 490)
+        zone_width_pct = zone_info.get("width_pct", 100)
+        zone_width_px = int(slide_width * zone_width_pct / 100 * 0.85)  # 패딩 제외
+
+        # 이 zone 안의 역할별 비중 비율 계산
+        total_weight = sum(info.get("weight", 0.25) for _, info in role_list)
+        if total_weight <= 0:
+            total_weight = 1.0
+
+        # 간격 제외
+        total_gap = gap_px * max(0, len(role_list) - 1)
+        available = zone_budget - total_gap
+
+        for role_name, role_info in role_list:
+            weight = role_info.get("weight", 0.25)
+            topic_ids = role_info.get("topic_ids", [])
+
+            # 비중 비율로 높이 할당
+            ratio = weight / total_weight
+            height_px = max(50, int(available * ratio))
+
+            # 블록 내부 제약 계산 — topic당 높이로 판단
+            topic_count = max(1, len(topic_ids))
+            per_topic_px = height_px // topic_count
+
+            # height_cost 허용 범위: topic당 높이 기준 (컨테이너 전체가 아님)
+            max_cost = _max_allowed_height_cost(per_topic_px)
+            font_size, padding, line_h = _determine_typography(height_px // topic_count)
+            constraints = _calculate_block_constraints(
+                height_px, zone_width_px, topic_count, font_size, line_h, padding
+            )
+            constraints["font_size_px"] = font_size
+            constraints["padding_px"] = padding
+            constraints["line_height"] = line_h
+
+            specs[role_name] = ContainerSpec(
+                role=role_name,
+                zone=zone_name,
+                topic_ids=topic_ids,
+                weight=weight,
+                height_px=height_px,
+                width_px=zone_width_px,
+                max_height_cost=max_cost,
+                block_constraints=constraints,
+            )
+
+    logger.info(
+        f"[O-1] 컨테이너 스펙: "
+        + ", ".join(f"{r}={s.height_px}px({s.max_height_cost})" for r, s in specs.items())
+    )
+    return specs
+
+
+def _max_allowed_height_cost(container_height_px: int) -> str:
+    """컨테이너 높이에서 허용되는 최대 height_cost."""
+    if container_height_px >= 350:
+        return "xlarge"
+    elif container_height_px >= 200:
+        return "large"
+    elif container_height_px >= 80:
+        return "medium"
+    else:
+        return "compact"
+
+
+def _determine_typography(per_block_height_px: int) -> tuple[float, int, float]:
+    """컨테이너 높이에 따른 폰트/패딩/줄간격 결정."""
+    if per_block_height_px >= 300:
+        return (15.2, 20, 1.7)
+    elif per_block_height_px >= 150:
+        return (14.0, 14, 1.6)
+    elif per_block_height_px >= 80:
+        return (13.0, 10, 1.5)
+    else:
+        return (12.0, 8, 1.4)
+
+
+def _calculate_block_constraints(
+    height_px: int,
+    width_px: int,
+    topic_count: int,
+    font_size_px: float,
+    line_height: float,
+    padding_px: int,
+) -> dict:
+    """컨테이너 크기에서 블록 내부 제약을 수학적으로 계산."""
+    per_topic_height = max(30, (height_px - padding_px * 2) // topic_count)
+    line_height_px = font_size_px * line_height
+    max_lines = max(1, int(per_topic_height / line_height_px))
+    chars_per_line = max(5, int((width_px - padding_px * 2) / (font_size_px * 0.95)))
+    max_items = max(1, max_lines // 2)
+    max_chars_total = max_lines * chars_per_line
+
+    return {
+        "max_lines": max_lines,
+        "max_items": max_items,
+        "chars_per_line": chars_per_line,
+        "max_chars_total": max(20, max_chars_total),
+        "max_chars_per_item": max(20, max_chars_total // max(1, max_items)),
+    }
+
+
+# ──────────────────────────────────────
+# O-1 유틸: topic_id → ContainerSpec 매핑
+# ──────────────────────────────────────
+def find_container_for_topic(
+    topic_id: int | None,
+    container_specs: dict[str, ContainerSpec],
+) -> ContainerSpec | None:
+    """topic_id로 해당 ContainerSpec을 찾는다."""
+    if topic_id is None:
+        return None
+    for spec in container_specs.values():
+        if topic_id in spec.topic_ids:
+            return spec
+    return None
+
+
+# ──────────────────────────────────────
+# O-3: 블록 스펙 확정
+# ──────────────────────────────────────
+def finalize_block_specs(
+    blocks: list[dict[str, Any]],
+    container_specs: dict[str, ContainerSpec],
+    catalog_map: dict[str, dict] | None = None,
+) -> list[dict[str, Any]]:
+    """각 블록의 내부 스펙을 컨테이너 크기에 맞게 확정한다.
+
+    결정론적. AI 호출 없음.
+
+    확정 필드:
+    - _container_height_px, _container_width_px
+    - _max_items, _max_chars_per_item, _max_chars_total
+    - _font_size_px, _padding_px, _line_height
+    """
+    for block in blocks:
+        tid = block.get("topic_id")
+        spec = find_container_for_topic(tid, container_specs)
+        if not spec:
+            continue
+
+        # 같은 컨테이너 안의 블록 수 (높이 분배)
+        siblings = [b for b in blocks
+                    if find_container_for_topic(b.get("topic_id"), container_specs) == spec
+                    and b.get("topic_id") is not None]
+        sibling_count = max(1, len(siblings))
+        per_block_height = max(40, spec.height_px // sibling_count)
+
+        # 폰트/패딩 결정
+        font_size, padding, line_h = _determine_typography(per_block_height)
+
+        # 블록별 제약 계산
+        constraints = _calculate_block_constraints(
+            per_block_height, spec.width_px, 1, font_size, line_h, padding
+        )
+
+        # 블록 타입별 세부 조정
+        block_type = block.get("type", "")
+        if block_type in ("dark-bullet-list",):
+            block["_max_items"] = min(constraints["max_items"], 5)
+            block["_max_chars_per_item"] = constraints["max_chars_per_item"]
+        elif block_type in ("card-numbered", "card-icon-desc"):
+            block["_max_items"] = constraints["max_items"]
+            block["_max_chars_per_item"] = constraints["max_chars_per_item"]
+        elif block_type in ("compare-2col-split", "compare-3col-badge", "table-simple-striped"):
+            block["_max_items"] = constraints["max_items"]  # 행 수
+            block["_max_chars_per_item"] = constraints["max_chars_per_item"]
+        elif block_type in ("comparison-2col",):
+            block["_max_chars_per_item"] = constraints["max_chars_total"] // 2
+        elif block_type in ("banner-gradient",):
+            block["_max_chars_total"] = constraints["chars_per_line"]
+        else:
+            pass  # 기본값 사용
+
+        # 공통 필드
+        block["_container_height_px"] = per_block_height
+        block["_container_width_px"] = spec.width_px
+        block["_max_chars_total"] = constraints["max_chars_total"]
+        block["_font_size_px"] = font_size
+        block["_padding_px"] = padding
+        block["_line_height"] = line_h
+
+    logger.info(
+        f"[O-3] 블록 스펙 확정: "
+        + ", ".join(
+            f"t{b.get('topic_id')}={b.get('_container_height_px','?')}px"
+            for b in blocks if b.get("topic_id") is not None
+        )
+    )
+    return blocks
+
+
+# ──────────────────────────────────────
+# 기존 유틸 (Phase L 호환)
+# ──────────────────────────────────────
+def calculate_trim_chars(
+    excess_px: int,
+    container_width_px: int,
+    font_size_px: float = DEFAULT_FONT_SIZE_PX,
+    line_height: float = DEFAULT_LINE_HEIGHT,
+    avg_char_width_px: float = DEFAULT_AVG_CHAR_WIDTH_PX,
+) -> int:
+    """초과 px에서 삭제할 글자 수를 계산한다."""
+    if excess_px <= 0:
+        return 0
+    line_height_px = font_size_px * line_height
+    lines_to_remove = math.ceil(excess_px / line_height_px)
+    chars_per_line = int(container_width_px / avg_char_width_px)
+    return max(lines_to_remove * chars_per_line, 10)