恢复GE版本为0.18，生成SNIPPET后自动生成rag_text流程

app/services/table_profiling.py

@@ -24,7 +24,6 @@ from app.services import LLMGateway
 from app.settings import DEFAULT_IMPORT_MODEL
 from app.services.import_analysis import (
     IMPORT_GATEWAY_BASE_URL,
-    build_import_gateway_headers,
     resolve_provider_from_model,
 )
 from app.utils.llm_usage import extract_usage as extract_llm_usage
@@ -533,7 +532,6 @@ async def _call_chat_completions(
     temperature: float = 0.2,
     timeout_seconds: Optional[float] = None,
 ) -> Any:
-    # Normalize model spec to provider+model and issue the unified chat call.
     provider, model_name = resolve_provider_from_model(model_spec)
     payload = {
         "provider": provider.value,
@@ -547,17 +545,16 @@ async def _call_chat_completions(
     payload_size_bytes = len(json.dumps(payload, ensure_ascii=False).encode("utf-8"))
 
     url = f"{IMPORT_GATEWAY_BASE_URL.rstrip('/')}/v1/chat/completions"
-    headers = build_import_gateway_headers()
     try:
+        # log the request whole info
         logger.info(
-            "Calling chat completions API %s with model=%s payload_size=%sB",
+            "Calling chat completions API %s with model %s and size %s and payload %s",
             url,
             model_name,
             payload_size_bytes,
+            payload,
         )
-        response = await client.post(
-            url, json=payload, timeout=timeout_seconds, headers=headers
-        )
+        response = await client.post(url, json=payload, timeout=timeout_seconds)
 
         response.raise_for_status()
     except httpx.HTTPError as exc:
@@ -706,7 +703,6 @@ async def _run_action_with_callback(
     input_payload: Any = None,
     model_spec: Optional[str] = None,
 ) -> Any:
-    # Execute a pipeline action and always emit a callback capturing success/failure.
     if input_payload is not None:
         logger.info(
             "Pipeline action %s input: %s",
@@ -789,6 +785,8 @@ async def process_table_profiling_job(
         "table_schema_version_id": request.table_schema_version_id,
         "llm_model": request.llm_model,
         "llm_timeout_seconds": timeout_seconds,
+        "workspace_id": request.workspace_id,
+        "rag_item_type": request.rag_item_type,
     }
 
     logging_request_payload = _profiling_request_for_log(request)

app/services/table_snippet.py

@@ -459,6 +459,18 @@ def _stable_rag_item_id(table_id: int, version_ts: int, snippet_id: str) -> int:
     return int(digest[:16], 16) % 9_000_000_000_000_000_000
 
 
+def _to_serializable(value: Any) -> Any:
+    if value is None or isinstance(value, (str, int, float, bool)):
+        return value
+    if isinstance(value, datetime):
+        return value.isoformat()
+    if isinstance(value, dict):
+        return {k: _to_serializable(v) for k, v in value.items()}
+    if isinstance(value, list):
+        return [_to_serializable(v) for v in value]
+    return str(value)
+
+
 def _build_rag_text(snippet: Dict[str, Any]) -> str:
     # Deterministic text concatenation for embedding input.
     parts: List[str] = []
@@ -512,7 +524,8 @@ def _prepare_rag_payloads(
             continue
         rag_item_id = _stable_rag_item_id(table_id, version_ts, snippet_id)
         rag_text = _build_rag_text(snippet)
-        merged_json = json.dumps(snippet, ensure_ascii=False)
+        serializable_snippet = _to_serializable(snippet)
+        merged_json = json.dumps(serializable_snippet, ensure_ascii=False)
         updated_at_raw = snippet.get("updated_at_from_action") or now
         if isinstance(updated_at_raw, str):
             try: