修复轻量拓展分析 NL 触发第一轮测试反馈（PM 裁断）

- P0 目录截断：_rough_candidates 关键词源扩展 extension_id token 与参数 flag；命中数降序排序；MAX_CATALOG_CHARS 12000→24000；截断时 reason 注明
- P1a 过滤缺失 extension_id 的记录，不再产出 None 候选
- P1b hmmsearch/hmmer/blast/blastp 不再全量触发，改为按关键词源匹配子集；仅「拓展分析/扩展分析/extension」允许全量候选
- P2 extension_name 改为 extension_id 紧凑形式（去尾缀时间戳截 30 字符）；message 中文模板（extension_id 为主、summary 括号补充）
- P2 nl-import-paths：目录同名冲突 _1/_2 后缀；多路径先整体校验后复制

apps/web/vue3/main.py

@@ -4451,7 +4451,8 @@ async def nl_import_paths(req: NlImportPathsRequest):
     """把服务器上的绝对路径（文件或目录）复制进当前 job 的 input/_nl_imports/。
 
     供轻量拓展分析自然语言触发向导使用；路径必须真实存在且解析后位于
-    /home/xzq/ 或 /data/ 之下。
+    /home/xzq/ 或 /data/ 之下。所有路径先整体校验，任一非法直接 400，
+    全部合法后才逐个复制。
     """
     job_id = str(req.job_id or "").strip()
     if not job_id:
@@ -4463,7 +4464,7 @@ async def nl_import_paths(req: NlImportPathsRequest):
         raise HTTPException(status_code=400, detail=f"无法创建 job 输入目录: {exc}") from exc
 
     allowed_roots = ("/home/xzq/", "/data/")
-    imported: dict[str, str] = {}
+    sources: list[Path] = []
     for raw_path in req.paths or []:
         raw_text = str(raw_path or "").strip()
         if not raw_text:
@@ -4483,8 +4484,12 @@ async def nl_import_paths(req: NlImportPathsRequest):
                 status_code=400,
                 detail=f"路径必须位于 {', '.join(allowed_roots)} 之下: {raw_text}",
             )
-        target_dir = job_dir / "_nl_imports"
-        target_dir.mkdir(parents=True, exist_ok=True)
+        sources.append(source)
+
+    target_dir = job_dir / "_nl_imports"
+    target_dir.mkdir(parents=True, exist_ok=True)
+    imported: dict[str, str] = {}
+    for source, raw_path in zip(sources, [str(p or "").strip() for p in req.paths or []]):
         if source.is_file():
             destination = target_dir / source.name
             candidate = destination
@@ -4495,16 +4500,21 @@ async def nl_import_paths(req: NlImportPathsRequest):
             try:
                 shutil.copy2(source, candidate)
             except OSError as exc:
-                raise HTTPException(status_code=400, detail=f"复制文件失败: {raw_text}: {exc}") from exc
+                raise HTTPException(status_code=400, detail=f"复制文件失败: {raw_path}: {exc}") from exc
             name = candidate.name
         else:
             destination = target_dir / source.name
+            candidate = destination
+            counter = 1
+            while candidate.exists():
+                candidate = target_dir / f"{source.name}_{counter}"
+                counter += 1
             try:
-                shutil.copytree(source, destination, dirs_exist_ok=True)
+                shutil.copytree(source, candidate, dirs_exist_ok=True)
             except OSError as exc:
-                raise HTTPException(status_code=400, detail=f"复制目录失败: {raw_text}: {exc}") from exc
-            name = destination.name
-        imported[raw_text] = f"input/{job_id}/_nl_imports/{name}"
+                raise HTTPException(status_code=400, detail=f"复制目录失败: {raw_path}: {exc}") from exc
+            name = candidate.name
+        imported[raw_path] = f"input/{job_id}/_nl_imports/{name}"
     return {"success": True, "imported": imported}
 
 
@@ -16553,6 +16563,7 @@ def _match_lightweight_extension_intent(message: str, job_id: str = "") -> dict[
 
     返回 dict 含 lightweight=True；任何失败返回 {}，绝不抛异常。
     """
+    records: list[dict] = []
     try:
         from gfa_core.extension_stage_lightweight import nl_intent
 
@@ -16571,10 +16582,20 @@ def _match_lightweight_extension_intent(message: str, job_id: str = "") -> dict[
         return {}
     if resolution is None or not resolution.extension_id:
         return {}
+    summary_hint = ""
+    for record in records:
+        if str(record.get("extension_id") or "") != resolution.extension_id:
+            continue
+        candidate = record.get("semantic_candidate")
+        if isinstance(candidate, dict):
+            first_sentence = re.split(r"[。.!?；;]", str(candidate.get("function_summary") or ""))[0].strip()
+            summary_hint = first_sentence[:60]
+        break
     params = {
         "extension_id": resolution.extension_id,
         "extensionName": resolution.extension_name or resolution.extension_id,
         "extension_store": "lightweight",
+        "function_summary": summary_hint,
         "llm_slots": {
             "input_overrides": resolution.input_overrides,
             "parameter_overrides": resolution.parameter_overrides,
@@ -16584,6 +16605,16 @@ def _match_lightweight_extension_intent(message: str, job_id: str = "") -> dict[
         "confidence": resolution.confidence,
         "parse_reason": resolution.reason,
     }
+    if summary_hint:
+        message_text = (
+            f"已识别为轻量拓展分析：{params['extension_id']}（{summary_hint}）。"
+            "请按引导上传或选择输入文件并核对参数后执行。"
+        )
+    else:
+        message_text = (
+            f"已识别为轻量拓展分析：{params['extension_id']}。"
+            "请按引导上传或选择输入文件并核对参数后执行。"
+        )
     return {
         "stage": "extension_stage",
         "stages": ["extension_stage"],
@@ -16596,7 +16627,7 @@ def _match_lightweight_extension_intent(message: str, job_id: str = "") -> dict[
         "taskType": "extension_stage",
         "lightweight": True,
         "confidence": resolution.confidence,
-        "message": f"已识别为轻量拓展分析：{params['extensionName']}。请按引导上传或选择输入文件并核对参数后执行。",
+        "message": message_text,
         "job_id": job_id,
         "workspace": str(PROJECT_ROOT / "workspace" / job_id) if job_id else "",
     }

gfa_core/extension_stage_lightweight/nl_intent.py

@@ -20,7 +20,7 @@ from pathlib import Path
 from typing import Any, Callable, Mapping
 
 PREFILL_THRESHOLD = 0.6
-MAX_CATALOG_CHARS = 12000
+MAX_CATALOG_CHARS = 24000
 MAX_JOB_FILE_HINTS = 200
 
 _ENGLISH_STOPWORDS = {"analysis", "data", "file", "run", "input", "output", "files", "using", "tool"}
@@ -105,19 +105,65 @@ def _summary_keywords(record: Mapping[str, Any]) -> tuple[list[str], list[str]]:
     return english, chinese
 
 
+def _extension_id_tokens(record: Mapping[str, Any]) -> list[str]:
+    """extension_id 里的英文 token（如 03___hmm_... → hmm、07_pangene__orthofinder_... → pangene/orthofinder）。"""
+    extension_id = str(record.get("extension_id") or "").lower()
+    return [
+        token for token in re.split(r"[^A-Za-z0-9]+", extension_id)
+        if len(token) >= 3 and token not in _ENGLISH_STOPWORDS
+    ]
+
+
+def _parameter_flag_tokens(record: Mapping[str, Any]) -> list[str]:
+    """parameters 的 flag 去前导 - 后作为关键词（如 --samples → samples）。"""
+    contract = record.get("runtime_binding_contract")
+    if not isinstance(contract, Mapping):
+        return []
+    tokens: list[str] = []
+    for item in contract.get("parameters") or []:
+        if not isinstance(item, Mapping):
+            continue
+        flag = str(item.get("flag") or "").strip().lower()
+        for token in re.split(r"[^A-Za-z0-9]+", flag):
+            if len(token) >= 3 and token not in _ENGLISH_STOPWORDS:
+                tokens.append(token)
+    return tokens
+
+
+def _record_keyword_sources(record: Mapping[str, Any]) -> list[str]:
+    """一条记录的全部粗筛关键词：summary 英文/中文子串 + extension_id token + 参数 flag。"""
+    english, chinese = _summary_keywords(record)
+    return english + chinese + _extension_id_tokens(record) + _parameter_flag_tokens(record)
+
+
 def _rough_candidates(message: str, records: list[dict]) -> list[str]:
-    """按关键词粗筛可能命中的记录 id；不含 LLM 调用。"""
+    """按关键词粗筛可能命中的记录 id；不含 LLM 调用。
+
+    返回按命中关键词数降序排列（同分按 id 排序）；「拓展分析/扩展分析/extension」
+    这类不含功能线索的强触发词才允许全部候选进目录，其余消息只返回命中子集。
+    """
     text = str(message or "").lower()
-    if any(token in text for token in ("hmmsearch", "hmmer", "blast", "blastp")):
-        return [str(record.get("extension_id")) for record in records if record.get("extension_id")]
     if any(token in text for token in ("拓展分析", "扩展分析", "extension")):
         return [str(record.get("extension_id")) for record in records if record.get("extension_id")]
-    matched: list[str] = []
+    scored: list[tuple[int, str]] = []
     for record in records:
-        english, chinese = _summary_keywords(record)
-        if any(token in text for token in english) or any(sub in text for sub in chinese):
-            matched.append(str(record.get("extension_id")))
-    return matched
+        extension_id = str(record.get("extension_id") or "").strip()
+        if not extension_id:
+            continue
+        hits = sum(1 for keyword in _record_keyword_sources(record) if keyword and keyword in text)
+        if hits:
+            scored.append((hits, extension_id))
+    scored.sort(key=lambda item: (-item[0], item[1]))
+    return [extension_id for _, extension_id in scored]
+
+
+def compact_extension_name(extension_id: str) -> str:
+    """extension_id 的紧凑展示形式：去掉尾缀时间戳（_后 10 位以上数字），再截断 30 字符。"""
+    value = str(extension_id or "").strip()
+    compact = re.sub(r"_\d{10,}$", "", value)
+    if compact:
+        return compact[:30]
+    return value[:30]
 
 
 def _rough_trigger(message: str, records: list[dict]) -> bool:
@@ -126,14 +172,19 @@ def _rough_trigger(message: str, records: list[dict]) -> bool:
 
 
 def _build_catalog(records: list[dict]) -> dict:
-    """构造发给 LLM 的拓展目录；序列化后总长超过 MAX_CATALOG_CHARS 时截断。"""
+    """构造发给 LLM 的拓展目录；序列化后总长超过 MAX_CATALOG_CHARS 时截断。
+
+    records 顺序即目录顺序（_rough_candidates 已按命中关键词数降序排好）。
+    """
     items: list[dict] = []
+    seen_ids: set[str] = set()
     total = 0
     truncated = False
     for record in records:
         extension_id = str(record.get("extension_id") or "").strip()
-        if not extension_id:
+        if not extension_id or extension_id in seen_ids:
             continue
+        seen_ids.add(extension_id)
         contract = record.get("runtime_binding_contract")
         if not isinstance(contract, Mapping):
             continue
@@ -279,9 +330,12 @@ def resolve_lightweight_extension_intent(
         return LightweightIntentResolution(None, None, {}, {}, None, 0.0, False, "no_published_records", [])
     if not candidates:
         return LightweightIntentResolution(None, None, {}, {}, None, 0.0, False, "no_trigger", [])
-    catalog = _build_catalog(
-        [record for record in published_records if str(record.get("extension_id") or "") in candidates]
+    candidate_ids = set(candidates)
+    ordered_records = sorted(
+        [record for record in published_records if str(record.get("extension_id") or "") in candidate_ids],
+        key=lambda record: candidates.index(str(record.get("extension_id") or "")),
     )
+    catalog = _build_catalog(ordered_records)
     if not catalog["extensions"]:
         return LightweightIntentResolution(None, None, {}, {}, None, 0.0, False, "no_trigger", candidates)
     hints = list(job_file_hints or [])[:MAX_JOB_FILE_HINTS]
@@ -322,14 +376,11 @@ def resolve_lightweight_extension_intent(
         )
 
     extension_id = sanitized["extension_id"]
-    selected = next(
-        (item for item in catalog["extensions"] if item.get("extension_id") == extension_id),
-        None,
-    )
-    extension_name = extension_id
-    if selected is not None:
-        first_sentence = re.split(r"[。.!?；;]", str(selected.get("function_summary") or ""))[0].strip()
-        extension_name = first_sentence[:60] or extension_id
+    # P2-4: extension_name 用 extension_id 的紧凑形式（去尾缀时间戳），不再用英文摘要句。
+    extension_name = compact_extension_name(extension_id)
+    reason = "ok"
+    if catalog.get("truncated"):
+        reason = "ok; 候选过多已截断，请补充工具名"
     return LightweightIntentResolution(
         extension_id=extension_id,
         extension_name=extension_name,
@@ -341,6 +392,6 @@ def resolve_lightweight_extension_intent(
             sanitized["confidence"] >= PREFILL_THRESHOLD
             and (sanitized["input_overrides"] or sanitized["parameter_overrides"])
         ),
-        reason="ok",
+        reason=reason,
         candidates_considered=candidates,
     )