feat(eval): memoir A/B chapter judging and eval-web parity with dialogue

- Judge baseline excerpt and library chapter separately; build_memoir_compare_summary for gate, nine-dim and leaf deltas.

- Memoir SSE chapter payload: baseline_judge, compare_summary, baseline_judge_error.

- MemoirJudgeOutput: loose score coercion and post-validate clamp; memoir judge prompt caps from settings.

- app-eval-web: two-column MemoirScoreCard layout, MemoirCompareSummary, chapter blocks and CSS.

- Add memoir_compare_summary, log_events, celery_log_context, memoir_pipeline_progress; tests and migration 0014.

- Misc: memory/evidence and enrichment paths, task/orchestrator updates, internal-eval docs, env examples.

api/app/core/agent_logging.py

@@ -1,28 +1,45 @@
 """
 Agent / LLM 诊断日志：耗时、输入输出规模、截断预览。
 
-- **详情**（完整 prompt 预览等）：仅在 ``LOG_LEVEL`` 为 ``TRACE`` / ``DEBUG`` 时通过 ``logger.debug`` 输出。
+- **详情**（prompt 预览 / hash / 响应预览）：仅在 ``LOG_LEVEL`` 为 ``TRACE`` / ``DEBUG`` 时通过 ``logger.debug`` 输出。
 - **摘要**（单行：耗时、字符数、operation 名）：当 ``LOG_AGENT_VERBOSE=1`` 时通过 ``logger.info`` 输出，
   便于生产环境在不把全局日志调到 DEBUG 的情况下排查 Agent 性能与路径。
 
-敏感内容：DEBUG 下会记录用户相关文本；``AGENT_LOG_MAX_CHARS=0`` 时记录全文，生产环境请勿长期开启 DEBUG。
+生产/预发建议 ``LOG_LEVEL=INFO``；需看 Agent 耗时与规模时可设 ``LOG_AGENT_VERBOSE=1``，无需长期 DEBUG。
+
+敏感内容：DEBUG 下会记录用户相关文本；``AGENT_LOG_MAX_CHARS=0`` 时预览不截断（完整输出，慎用）。
 
 配置（节选）：``AGENT_LOG_OMIT_SYSTEM_MESSAGE_BODY``（默认 true）省略聊天 System 正文，仅打 len+sha12；
 ``AGENT_LOG_JSON_PROMPT_PREFIX_CHARS`` + ``AGENT_LOG_JSON_PROMPT_PREFIX_ONLY_IF_LEN_GT`` 在 DEBUG 下跳过
-超长单段 prompt 的前缀再预览。
+超长单段 ``*.prompt`` 的前缀再预览；
+``AGENT_LOG_PROMPT_MODE=hash_only`` 时 ``*.prompt`` 仅输出 sha12 + 长度，无正文；
+``AGENT_LOG_PROMPT_DEDUP=1`` 时同一 label 连续相同全文则跳过重复行。
 """
 
 from __future__ import annotations
 
+import hashlib
+import threading
 import time
 from contextlib import contextmanager
 from typing import Any, Iterator
 
 from app.core.config import settings
 
+_dedup_lock = threading.Lock()
+_last_prompt_sha256_by_label: dict[str, str] = {}
+
+
+def _payload_sha256_hex(text: str) -> str:
+    return hashlib.sha256(text.encode("utf-8", errors="replace")).hexdigest()
+
+
+def _payload_sha12(text: str) -> str:
+    return _payload_sha256_hex(text)[:12]
+
 
 def agent_verbose_enabled() -> bool:
-    """是否输出含完整 prompt 预览等 DEBUG 级详情。"""
+    """是否输出含 prompt/response 等 DEBUG 级详情。"""
     raw = (settings.log_level or "INFO").strip().upper()
     return raw in ("TRACE", "DEBUG")
 
@@ -97,26 +114,55 @@ def log_agent_payload(
     *,
     max_chars: int | None = None,
 ) -> None:
-    """在 DEBUG 下记录文本长度与截断预览。"""
+    """在 DEBUG 下记录文本长度与截断预览（*.prompt 可 hash_only / 去重）。"""
     if not agent_verbose_enabled():
         return
     raw = text or ""
     total_len = len(raw)
+    digest = _payload_sha256_hex(raw)
+    sha12 = digest[:12]
+
+    is_prompt = label.endswith(".prompt")
+    if is_prompt and settings.agent_log_prompt_dedup:
+        with _dedup_lock:
+            if _last_prompt_sha256_by_label.get(label) == digest:
+                logger.debug(
+                    "agent_payload_skipped label={} reason=same_as_previous sha12={} total_len={}",
+                    label,
+                    sha12,
+                    total_len,
+                )
+                return
+            _last_prompt_sha256_by_label[label] = digest
+
     preview_source = raw
     extra_note = ""
     if (
-        label.endswith(".prompt")
+        is_prompt
         and settings.agent_log_json_prompt_prefix_chars > 0
         and total_len > settings.agent_log_json_prompt_prefix_only_if_len_gt
     ):
         skip = settings.agent_log_json_prompt_prefix_chars
         preview_source = raw[skip:]
         extra_note = f" skipped_prefix_chars={skip}"
+
+    mode = (settings.agent_log_prompt_mode or "preview").strip().lower()
+    if is_prompt and mode == "hash_only":
+        logger.debug(
+            "agent_payload label={} total_len={} sha12={} mode=hash_only{}",
+            label,
+            total_len,
+            sha12,
+            extra_note,
+        )
+        return
+
     preview = truncate_for_log(preview_source, max_chars=max_chars)
     logger.debug(
-        "agent_payload {} total_len={}{} preview={}",
+        "agent_payload {} total_len={}{} sha12={} preview={}",
         label,
         total_len,
         extra_note,
+        sha12,
         preview,
     )