From 1a1b5aa53e4eaa53e3cc50f6c7b88c24be7e6866 Mon Sep 17 00:00:00 2001
From: NeedmeFordev <124189514+spider-yamet@users.noreply.github.com>
Date: Tue, 14 Apr 2026 04:55:20 -0700
Subject: [PATCH] Fix: respect the internet toggle before running Tavily web
 search (#14051) (#14052)

### What problem does this PR solve?

Fixes #14051.

The chat UI already sends an `internet` flag with each request, but the
backend previously triggered Tavily web retrieval whenever
`prompt_config.tavily_api_key` was configured. As a result, web search
could still run even when the internet toggle was off.

This PR makes web search an explicit opt-in at request time:
- `tavily_api_key` only indicates that web search is available
- Tavily retrieval runs only when `internet` is explicitly enabled
- the same behavior now applies to both the normal retrieval path and
the deep-research / reasoning path

This also fixes the no-KB fallback case so chats without KBs fall back
to normal solo chat when `internet` is off.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
---
 api/db/services/dialog_service.py             | 28 +++++++++++++++++--
 ...tructured_query_decomposition_retrieval.py |  6 ++--
 2 files changed, 30 insertions(+), 4 deletions(-)

diff --git a/api/db/services/dialog_service.py b/api/db/services/dialog_service.py
index 83f79c285a..ede9bb0019 100644
--- a/api/db/services/dialog_service.py
+++ b/api/db/services/dialog_service.py
@@ -49,6 +49,27 @@ from common.string_utils import remove_redundant_spaces
 from common import settings
 
 
+def _normalize_internet_flag(value):
+    if isinstance(value, bool):
+        return value
+    if isinstance(value, (int, float)) and value in (0, 1):
+        return bool(value)
+    if isinstance(value, str):
+        normalized = value.strip().lower()
+        if normalized in {"true", "1", "yes", "on"}:
+            return True
+        if normalized in {"false", "0", "no", "off", ""}:
+            return False
+    return None
+
+
+def _should_use_web_search(prompt_config, internet=None):
+    if not prompt_config.get("tavily_api_key"):
+        return False
+    normalized = _normalize_internet_flag(internet)
+    return normalized is True
+
+
 class DialogService(CommonService):
     model = Dialog
 
@@ -461,7 +482,9 @@ def repair_bad_citation_formats(answer: str, kbinfos: dict, idx: set):
 async def async_chat(dialog, messages, stream=True, **kwargs):
     logging.debug("Begin async_chat")
     assert messages[-1]["role"] == "user", "The last content of this conversation is not from user."
-    if not dialog.kb_ids and not dialog.prompt_config.get("tavily_api_key"):
+    use_web_search = _should_use_web_search(dialog.prompt_config, kwargs.get("internet"))
+    logging.debug("web_search kb=%s tavily=%s internet=%r enabled=%s", bool(dialog.kb_ids), bool(dialog.prompt_config.get("tavily_api_key")), kwargs.get("internet"), use_web_search)
+    if not dialog.kb_ids and not use_web_search:
         async for ans in async_chat_solo(dialog, messages, stream):
             yield ans
         return
@@ -587,6 +610,7 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
                     vector_similarity_weight=0.3,
                     doc_ids=attachments,
                 ),
+                internet_enabled=use_web_search,
             )
             queue = asyncio.Queue()
             async def callback(msg:str):
@@ -629,7 +653,7 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
                     if cks:
                         kbinfos["chunks"] = cks
                 kbinfos["chunks"] = retriever.retrieval_by_children(kbinfos["chunks"], tenant_ids)
-            if prompt_config.get("tavily_api_key"):
+            if use_web_search:
                 tav = Tavily(prompt_config["tavily_api_key"])
                 tav_res = tav.retrieve_chunks(" ".join(questions))
                 kbinfos["chunks"].extend(tav_res["chunks"])
diff --git a/rag/advanced_rag/tree_structured_query_decomposition_retrieval.py b/rag/advanced_rag/tree_structured_query_decomposition_retrieval.py
index 214485c3b0..11af6aa46b 100644
--- a/rag/advanced_rag/tree_structured_query_decomposition_retrieval.py
+++ b/rag/advanced_rag/tree_structured_query_decomposition_retrieval.py
@@ -28,12 +28,14 @@ class TreeStructuredQueryDecompositionRetrieval:
                  chat_mdl: LLMBundle,
                  prompt_config: dict,
                  kb_retrieve: partial = None,
-                 kg_retrieve: partial = None
+                 kg_retrieve: partial = None,
+                 internet_enabled: bool = False,
                  ):
         self.chat_mdl = chat_mdl
         self.prompt_config = prompt_config
         self._kb_retrieve = kb_retrieve
         self._kg_retrieve = kg_retrieve
+        self.internet_enabled = internet_enabled
         self._lock = asyncio.Lock()
 
     async def _retrieve_information(self, search_query):
@@ -47,7 +49,7 @@ class TreeStructuredQueryDecompositionRetrieval:
 
         # 2. Web retrieval (if Tavily API is configured)
         try:
-            if self.prompt_config.get("tavily_api_key"):
+            if self.internet_enabled and self.prompt_config.get("tavily_api_key"):
                 tav = Tavily(self.prompt_config["tavily_api_key"])
                 tav_res = tav.retrieve_chunks(search_query)
                 kbinfos["chunks"].extend(tav_res["chunks"])