@easysociable/cli 0.1.4 → 0.1.6

package/content-engine/starter/topics/_template-item.md

@@ -18,55 +18,58 @@ product_fit: none
 recommendation_ids: none
 ---
 
-# 选题判断卡
+# Decision card
 
-> 这是一条准备讨论的内容机会，不是最终发布标题。
+> A content opportunity under discussion — not a final publish title.
 
-**结论：** ready | backlog | parked  
-**状态原因：** 一句话说明为什么得到这个结论  
-**适合谁：** 一句话说明具体受众和场景  
-**核心判断：** 一句话说明这条内容真正要讨论什么  
-**下一步：** 开始生成 / 补需求 / 补证据 / 修改角度 / 放弃
-**产品连接：** none | P-… · direct | adjacent · 一句话说明连接方式
-**生成路径：** demand | demand_to_offer | demand_to_review | review | offer_education | profile_thesis
+**Language:** Match `profiles/<id>/voice.md` **Language** (else pillars “Primary public language”, else `engine.json` locale). Readable judgment + `one_liner` must use that language. Verbatim Needs/Capture quotes stay untranslated. Operator chat may add a short gloss in another language; the file does not.
 
-## 这条内容要解决什么问题？
+**Verdict:** ready | backlog | parked  
+**Why this status:** one sentence  
+**Who it’s for:** one sentence — concrete audience and situation  
+**Core judgment:** one sentence — what this piece actually argues  
+**Next step:** open a run / fill needs / fill evidence / re-angle / kill  
+**Product link:** none | P-… · direct | adjacent · one sentence  
+**Generation path:** demand | demand_to_offer | demand_to_review | review | offer_education | profile_thesis  
+*(Agent must ask the user to choose before `produce`; do not silent-default.)*
 
-用 2–4 句话说明现实问题。不要复述标题。
+## What problem does this solve?
 
-## 他们现在相信什么？
+2–4 sentences on the real-world problem. Do not restate a title.
 
-写出受众当前的错误判断或困惑。
+## What do they believe now?
 
-## 我们准备讲哪一刀？
+Name the false belief or stuck confusion.
 
-说明我们的独特角度，不写完整标题，也不要把来源内容改写成我们的观点。
+## What cut are we making?
 
-## 为什么现在值得讲？
+Our angle only. Not a full headline. Do not rewrite a source post as our thesis.
 
-说明发生了什么变化，以及为什么当前受众需要面对它。不是新闻摘要。
+## Why now?
 
-## 看完后应该改变什么？
+What changed, and why this audience must face it now. Not a news summary.
 
-用“之前 → 之后”表达用户认知或行动上的变化。
+## What should change after reading?
 
-## 凭什么这么判断？
+Before → after for belief or action.
 
-用人话列出 Needs、Product、Recommendation、Capture 或其他依据。
+## What is this based on?
 
-## 当前缺口
+Plain-language Needs, Product, Recommendation, Capture, or other evidence.
 
-还缺什么需求证据、案例、用户反馈或产品事实。没有缺口就写“无”。
+## Current gaps
 
-## 建议形式
+Missing need evidence, cases, feedback, or product facts. Write `none` if empty.
 
-平台、Lane 和推荐内容形式。暂时不写完整脚本。
+## Suggested form
+
+Platform, lane, and form. No full script yet.
 
 ---
 
-## 评分与追踪附录
+## Score / trace appendix
 
-以下内容服务于复核、机器读取和后续交接，不是主阅读面。
+For review, machines, and handoff — not the primary reading surface.
 
 ### Score
 

package/content-engine/starter/topics/scoring.md

@@ -185,7 +185,9 @@ raise the audience score.
 
 ### Generation-mode gates
 
-Before scoring, set the smallest honest `generation_mode`:
+Before scoring a `produce` Topic, **ask the user to choose `generation_mode`**.
+Recommend `demand` when Need-led; do not silently assume. Then enforce required
+inputs:
 
 | mode | Required |
 |------|----------|
@@ -198,7 +200,8 @@ Before scoring, set the smallest honest `generation_mode`:
 
 Do not use `review` when the only evidence is vendor marketing. Review entries
 must state limitations, research date, and commercial disclosure where
-relevant. Product-only and recommendation-only inputs do not imply demand.
+relevant. Product-only and recommendation-only inputs do not imply demand and
+never raise audience / `ready` alone.
 
 When `evidence < 5`, agent should flag an **evidence gap** on the topic item and may suggest **research** before finalizing the score. Topics can be re-scored after research fills the gap.
 

package/content-engine/workbench/build.py

@@ -108,8 +108,13 @@ def extract_tables(md: str) -> list[dict]:
     blocks: list[list[str]] = []
     cur: list[str] = []
     for line in md.splitlines():
-        if line.strip().startswith("|"):
+        stripped = line.strip()
+        if stripped.startswith("|"):
             cur.append(line)
+        elif not stripped:
+            # Blank lines inside a GFM table must not split the block,
+            # or later rows are parsed as a new headerless table.
+            continue
         elif cur:
             blocks.append(cur)
             cur = []
@@ -379,7 +384,7 @@ def _parse_compact_count(raw: str) -> str:
 def _fill_engagement_from_blob(out: dict[str, str], blob: str) -> None:
     if not blob:
         return
-    # Prefer kind-first matches (discovery + list form), then num-first (capture prose).
+    # Prefer kind-first matches (list form), then num-first (capture prose).
     for kind, num in ENGAGEMENT_KIND_NUM_RE.findall(blob):
         key = f"src_{_normalize_metric_kind(kind)}"
         if key in out and not out[key]:
@@ -395,10 +400,10 @@ def _fill_engagement_from_blob(out: dict[str, str], blob: str) -> None:
 
 
 def parse_engagement_snapshot(text: str) -> dict[str, str]:
-    """Parse engagement into src_* fields from Discovery or Capture/Raw notes.
+    """Parse engagement into src_* fields from Capture / Raw notes.
 
     Supports:
-    - Discovery: `engagement snapshot: likes 113 / replies 18 / ... @ 2026-08-30`
+    - Compact: `engagement snapshot: likes 113 / replies 18 / ... @ 2026-08-30`
     - Capture prose: `Engagement snapshot (fetch DATE): ~92.9k views · 1086 likes · ...`
     - Raw list under `## Engagement snapshot` with `- likes: 3`
     Parenthetical secondary demos after the primary snapshot are ignored.
@@ -446,7 +451,7 @@ def build_graph() -> dict:
     for p in profiles:
         base = p["path"].rstrip("/") + "/"
         files = {}
-        for name in ("README", "voice", "audience", "pillars", "lanes"):
+        for name in ("README", "voice", "audience", "pillars", "lanes", "keywords", "handles"):
             rel = f"{base}{name}.md"
             if (ROOT / rel).exists():
                 files[name.lower()] = rel
@@ -562,6 +567,174 @@ def build_graph() -> dict:
         )
         # Topic edges are wired after topic nodes exist (see post-pass below).
 
+    hits_path = ROOT / "hits" / "_index.md"
+    if hits_path.exists():
+        hits_rows = table_with(
+            hits_path.read_text(encoding="utf-8"),
+            "id",
+            "date",
+            "profile",
+            "platform",
+            "keyword_id",
+            "source",
+            "hit_kind",
+            "signal",
+            "status",
+            "excerpt",
+            "src_likes",
+            "src_replies",
+            "src_bookmarks",
+            "src_views",
+            "engagement_at",
+            "file",
+        )
+        if not hits_rows:
+            hits_rows = table_with(
+                hits_path.read_text(encoding="utf-8"),
+                "id",
+                "date",
+                "profile",
+                "platform",
+                "source",
+                "hit_kind",
+                "signal",
+                "status",
+                "excerpt",
+                "file",
+            )
+        for row in hits_rows:
+            ident = row["id"]
+            if not ident.startswith("H-"):
+                continue
+            fpath = wiki_path(row["file"]) or f"hits/entries/{ident}.md"
+            if not fpath.endswith(".md"):
+                fpath = f"{fpath}.md"
+            excerpt = (row.get("excerpt") or "").strip()
+            keyword_id = (row.get("keyword_id") or "").strip()
+            if not keyword_id:
+                m = re.search(r"\bK-\d{2}\b", row.get("source", "") or "")
+                if m:
+                    keyword_id = m.group(0)
+            eng = {
+                "src_likes": row.get("src_likes", "") or "",
+                "src_replies": row.get("src_replies", "") or "",
+                "src_bookmarks": row.get("src_bookmarks", "") or "",
+                "src_views": row.get("src_views", "") or "",
+                "engagement_at": row.get("engagement_at", "") or "",
+            }
+            source_url = ""
+            if fpath and (ROOT / fpath).exists():
+                body = read(fpath)
+                eng = merge_engagement(eng, parse_engagement_snapshot(body))
+                fm = re.match(r"^---\n(.*?)\n---", body, re.S)
+                if fm:
+                    murl = re.search(r"^source_url:\s*(\S+)", fm.group(1), re.M)
+                    if murl:
+                        source_url = murl.group(1).strip()
+                    if not keyword_id:
+                        mk = re.search(r"^keyword_id:\s*(\S+)", fm.group(1), re.M)
+                        if mk:
+                            keyword_id = mk.group(1).strip()
+            add_node(
+                node(
+                    "hit",
+                    ident,
+                    excerpt or f"Hit {ident}",
+                    fpath,
+                    {
+                        "date": row["date"],
+                        "profile": row["profile"],
+                        "platform": row.get("platform", "") or "",
+                        "keyword_id": keyword_id,
+                        "source": row.get("source", "") or keyword_id,
+                        "hit_kind": row.get("hit_kind", "") or "",
+                        "signal": row.get("signal", "") or "",
+                        "status": row["status"],
+                        "excerpt": excerpt,
+                        "source_url": source_url,
+                        "src_likes": eng.get("src_likes", "") or "",
+                        "src_replies": eng.get("src_replies", "") or "",
+                        "src_bookmarks": eng.get("src_bookmarks", "") or "",
+                        "src_views": eng.get("src_views", "") or "",
+                        "engagement_at": eng.get("engagement_at", "") or "",
+                        "src_heat_note": "来源帖热度，不是你的帖",
+                    },
+                )
+            )
+
+    # Keyword recipes (profile listening config) + hit counts for Hits rail
+    keywords: list[dict] = []
+    for p in profiles:
+        kpath = ROOT / p["path"].rstrip("/") / "keywords.md"
+        if not kpath.exists():
+            continue
+        krows = table_with(
+            kpath.read_text(encoding="utf-8"),
+            "id",
+            "track",
+            "query",
+            "platform",
+            "intent",
+            "layer",
+            "pillar",
+            "why",
+            "weight",
+            "status",
+            "seed_from",
+            "pair_id",
+            "volume",
+            "kd",
+            "serp_intent",
+            "volume_source",
+            "volume_at",
+            "ask_engines",
+            "last_scanned",
+        )
+        if not krows:
+            krows = table_with(
+                kpath.read_text(encoding="utf-8"),
+                "id",
+                "query",
+                "platform",
+                "intent",
+                "layer",
+                "pillar",
+                "why",
+                "weight",
+                "status",
+                "last_scanned",
+            )
+        for row in krows:
+            kid = (row.get("id") or "").strip()
+            if not kid.startswith("K-"):
+                continue
+            track = (row.get("track") or "").strip() or "listen"
+            keywords.append(
+                {
+                    "id": kid,
+                    "profile": p["id"],
+                    "track": track,
+                    "query": row.get("query", "") or "",
+                    "platform": row.get("platform", "") or "",
+                    "intent": row.get("intent", "") or "",
+                    "layer": row.get("layer", "") or "",
+                    "pillar": row.get("pillar", "") or "",
+                    "why": row.get("why", "") or "",
+                    "weight": row.get("weight", "") or "",
+                    "status": row.get("status", "") or "",
+                    "seed_from": row.get("seed_from", "") or "",
+                    "pair_id": row.get("pair_id", "") or "",
+                    "volume": row.get("volume", "") or "",
+                    "kd": row.get("kd", "") or "",
+                    "serp_intent": row.get("serp_intent", "") or "",
+                    "volume_source": row.get("volume_source", "") or "",
+                    "volume_at": row.get("volume_at", "") or "",
+                    "ask_engines": row.get("ask_engines", "") or "",
+                    "last_scanned": row.get("last_scanned", "") or "",
+                    "path": f"{p['path'].rstrip('/')}/keywords.md",
+                }
+            )
+
     products_path = ROOT / "products" / "_index.md"
     if products_path.exists():
         product_rows = table_with(
@@ -684,6 +857,7 @@ def build_graph() -> dict:
         "need_ids",
         "product_ids",
         "product_fit",
+        "recommendation_ids",
         "run",
     )
     if topic_rows:
@@ -801,6 +975,7 @@ def build_graph() -> dict:
                     "need_ids": row.get("need_ids", "") or "",
                     "product_ids": row.get("product_ids", "") or "",
                     "product_fit": row.get("product_fit", "") or "",
+                    "recommendation_ids": row.get("recommendation_ids", "") or "",
                     "run_id": row["run"].strip(),
                     "usage": usage,
                     "src_heat_note": "来源帖热度，不是你的帖",
@@ -811,6 +986,13 @@ def build_graph() -> dict:
             edge(nid("topic", ident), nid("capture", row["capture_id"]), "from_capture")
         if row["run"].strip().startswith("RUN-"):
             edge(nid("topic", ident), nid("run", row["run"].strip()), "opened_run")
+        # Index cells are enough for edges; item frontmatter may repeat the same ids.
+        for need_id in re.findall(r"\bN-\d{8}-\d{2}\b", row.get("need_ids", "") or ""):
+            edge(nid("topic", ident), nid("need", need_id), "from_need")
+        for product_id in PRODUCT_RE.findall(row.get("product_ids", "") or ""):
+            edge(nid("topic", ident), nid("product", product_id), "uses_product")
+        for recommendation_id in RECOMMENDATION_RE.findall(row.get("recommendation_ids", "") or ""):
+            edge(nid("topic", ident), nid("recommendation", recommendation_id), "reviews")
         if item and (ROOT / item).exists():
             body = read(item)
             fm = re.match(r"^---\n(.*?)\n---", body, re.S)
@@ -1231,14 +1413,16 @@ def build_graph() -> dict:
         "pillars_by_profile": pillars_by_profile,
         "pillars": flat_pillars,
         "needs_index_present": (ROOT / "needs" / "_index.md").exists(),
+        "hits_index_present": (ROOT / "hits" / "_index.md").exists(),
         "products_index_present": (ROOT / "products" / "_index.md").exists(),
+        "keywords": keywords,
         "nodes": list(nodes.values()),
         "edges": uniq,
     }
 
 
 def _self_check() -> None:
-    # ponytail: tiny parser asserts; expand if capture/discovery formats drift again.
+    # ponytail: tiny parser asserts; expand if capture/raw formats drift again.
     d = parse_engagement_snapshot(
         "- **engagement snapshot:** likes 113 / replies 18 / bookmarks 17 / views 9574 @ 2026-08-30\n"
     )