feat(chat): add Querit web search provider (#17813)

This commit is contained in:
EthanZhang
2026-08-05 09:54:46 +08:00
committed by GitHub
parent 4d68e154ce
commit bdcd8aadde
32 changed files with 1253 additions and 134 deletions

View File

@@ -55,7 +55,7 @@ from rag.prompts.generator import (
sufficiency_select,
)
from api.db.db_models import Document, Knowledgebase
from rag.utils.tavily_conn import Tavily
from rag.utils.web_search_conn import WebSearchProvider
# Tokens held back from the model's context when fitting retrieved evidence
@@ -75,7 +75,7 @@ class RAGTools:
embed_mdl: LLMBundle | None = None,
kb_ids: List[str] | None = None,
kbs: list[Knowledgebase] | None = None,
tav: Tavily | None = None,
web_search: WebSearchProvider | None = None,
meta_data_filter: dict | None = None,
doc_scope: List[str] | None = None,
user_defined_prompts: dict | None = None,
@@ -106,7 +106,7 @@ class RAGTools:
for kb in kbs:
_exclude_sql_kb(kb)
self.tav = tav
self.web_search = web_search
self.meta_data_filter = meta_data_filter
self.doc_scope = list(dict.fromkeys(doc_scope)) if doc_scope is not None else None
self.user_defined_prompts = user_defined_prompts or {}
@@ -137,7 +137,7 @@ class RAGTools:
return bool(self.sql_kbs and self.field_map)
def has_web(self) -> bool:
return self.tav is not None
return self.web_search is not None
def has_llm(self) -> bool:
return self.chat_mdl is not None
@@ -423,15 +423,15 @@ class RAGTools:
return {"chunks": kbinfos.get("chunks", []), "doc_aggs": kbinfos.get("doc_aggs", [])}
async def web_retrieve(self, query: str) -> dict[str, list]:
"""Retrieve chunks from the public web (Tavily). Raw kbinfos shape."""
if self.tav is None:
"""Retrieve chunks from the public web. Raw kbinfos shape."""
if self.web_search is None:
return {"chunks": [], "doc_aggs": []}
try:
tav_res = await thread_pool_exec(self.tav.retrieve_chunks, query)
web_res = await thread_pool_exec(self.web_search.retrieve_chunks, query)
except Exception:
logging.exception("web_retrieve failed")
return {"chunks": [], "doc_aggs": []}
return {"chunks": tav_res.get("chunks", []), "doc_aggs": tav_res.get("doc_aggs", [])}
return {"chunks": web_res.get("chunks", []), "doc_aggs": web_res.get("doc_aggs", [])}
async def structured_retrieve(self, question: str) -> dict[str, Any]:
"""Query the structured (tabular) KBs by translating to SQL.

View File

@@ -768,8 +768,8 @@ async def web_search(tools, query: str, keywords: str = "") -> dict:
from common.misc_utils import thread_pool_exec
effective_query = f"{query} {keywords}".strip() if keywords else query
tav_res = await thread_pool_exec(tools.tav.retrieve_chunks, effective_query)
return {"chunks": tav_res.get("chunks", []), "doc_aggs": tav_res.get("doc_aggs", [])}
web_res = await thread_pool_exec(tools.web_search.retrieve_chunks, effective_query)
return {"chunks": web_res.get("chunks", []), "doc_aggs": web_res.get("doc_aggs", [])}
except Exception:
_LOG.exception("web_search failed")
return {"chunks": [], "doc_aggs": []}

View File

@@ -19,7 +19,7 @@ from functools import partial
from api.db.services.llm_service import LLMBundle
from rag.prompts import kb_prompt
from rag.prompts.generator import sufficiency_check, multi_queries_gen
from rag.utils.tavily_conn import Tavily
from rag.utils.web_search_conn import create_web_search_provider
from timeit import default_timer as timer
@@ -49,13 +49,13 @@ class TreeStructuredQueryDecompositionRetrieval:
except Exception as e:
logging.error(f"Knowledge base retrieval error: {e}")
# 2. Web retrieval (if Tavily API is configured)
# 2. Web retrieval (if a web search provider is configured)
try:
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"])
kbinfos["doc_aggs"].extend(tav_res["doc_aggs"])
web_search = create_web_search_provider(self.prompt_config) if self.internet_enabled else None
if web_search:
web_res = web_search.retrieve_chunks(search_query)
kbinfos["chunks"].extend(web_res["chunks"])
kbinfos["doc_aggs"].extend(web_res["doc_aggs"])
except Exception as e:
logging.error(f"Web retrieval error: {e}")