@roll-agent/octopus-agent 0.0.2 → 0.0.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@roll-agent/octopus-agent",
-  "version": "0.0.2",
+  "version": "0.0.4",
   "description": "Octopus Sponge NL2SQL Agent",
   "license": "UNLICENSED",
   "keywords": [

package/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "octopus-skill"
-version = "0.0.2"
+version = "0.0.4"
 description = "Octopus Sponge NL2SQL Agent"
 readme = "README.md"
 requires-python = ">=3.11"

package/src/octopus_sponge_skill/_version.py

@@ -1,3 +1,3 @@
 from __future__ import annotations
 
-__version__ = "0.0.2"
+__version__ = "0.0.4"

package/src/octopus_sponge_skill/agent.py

@@ -95,6 +95,9 @@ class SpongeAgent:
         )
         sql_question = _strip_result_format_instruction(user_question)
         schema = self._ensure_schema()
+        business_clarification = _business_metric_clarification(sql_question, schema)
+        if business_clarification is not None:
+            return _business_metric_clarification_result(business_clarification)
         schema_version = str(schema.get("schemaVersion")) if schema.get("schemaVersion") is not None else None
         repaired_sql: str | None = None
         prefer_rule_based = is_recruiting_match_effect_query(sql_question)
@@ -439,6 +442,16 @@ def _query_only_result() -> QueryResult:
     )
 
 
+def _business_metric_clarification_result(answer: str) -> QueryResult:
+    return QueryResult(
+        answer=answer,
+        generated_sql="",
+        normalized_sql="",
+        repaired_sql=None,
+        validation={},
+    )
+
+
 def _generate_with_sampler(
     *,
     generator: SamplingSqlGenerator,
@@ -530,6 +543,71 @@ def _is_mutation_analytics_request(question: str) -> bool:
     )
 
 
+def _business_metric_clarification(question: str, schema: dict[str, Any]) -> str | None:
+    if is_recruiting_match_effect_query(question):
+        return None
+    metric_terms = _ambiguous_business_metric_terms(question)
+    if not metric_terms:
+        return None
+    if _schema_has_metric_definition(question, schema, metric_terms):
+        return None
+
+    lines = [
+        "这个查询涉及业务口径，我暂时不直接生成 SQL，避免结果口径不准。",
+        "",
+        "请先确认：",
+        f"1. 指标定义：“{'、'.join(metric_terms)}”具体包含哪些数据？",
+        "2. 时间口径：按哪个时间字段统计？",
+        "3. 状态口径：哪些状态算有效/成功/完成/异常？",
+        "4. 聚合维度：需要按什么维度展示？",
+    ]
+    return "\n".join(lines)
+
+
+def _ambiguous_business_metric_terms(question: str) -> list[str]:
+    terms = (
+        "漏斗",
+        "转化率",
+        "转化",
+        "有效",
+        "活跃",
+        "成功率",
+        "留存",
+        "履约",
+        "异常",
+        "质量",
+        "效率",
+        "趋势",
+        "健康度",
+        "达成率",
+        "完成率",
+    )
+    return [term for term in terms if term in question]
+
+
+def _schema_has_metric_definition(question: str, schema: dict[str, Any], metric_terms: list[str]) -> bool:
+    metric_texts: list[str] = []
+    for metric in schema.get("metrics") or []:
+        if not isinstance(metric, dict):
+            continue
+        parts: list[str] = []
+        for key in ("name", "term", "displayName", "definition", "description", "aggregation"):
+            value = metric.get(key)
+            if isinstance(value, str):
+                parts.append(value)
+        filters = metric.get("filters")
+        if isinstance(filters, list):
+            parts.extend(str(item) for item in filters)
+        metric_text = " ".join(parts)
+        if metric_text:
+            metric_texts.append(metric_text)
+            name = metric.get("name")
+            if isinstance(name, str) and name and name in question:
+                return True
+
+    return any(term in metric_text for term in metric_terms for metric_text in metric_texts)
+
+
 def _apply_default_limit(sql: str, question: str, default_limit: int) -> str:
     if _question_requests_limit(question):
         return sql