Avoid external calls?

[markod0925]

This tool seems to be intended to fit into the "smart" web search processes and AI chat consultation of your documents. For this purpose, local resources (ex. Ollama) can be used that should guarantee absolute privacy to the user.

Given this hypothesis, I find it counterproductive that external calls can occur without explicit warning to the user, among other things to servers that are sometimes known (but too general),

[05/02/25 09:39:52.576] INFO [8368-19144] Checking dense embedder for KB graphrag ... (C:\Users\XXXXX\AppData\Local\Programs\Python\Python312\Lib\site-packages\
leettools\cli\cli_utils.py:134 setup_org_kb_user)
Exception reading Public Suffix List url https://publicsuffix.org/list/public_suffix_list.dat
urllib3.exceptions.SSLError: [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: self-signed certificate in certificate chain (_ssl.c:1000)

sometimes non-standard

[05/02/25 09:42:42.347] ERROR Error in getting answer for query: HTTPSConnectionPool(host='svc.leettools.com', port=8098): Max retries exceeded with url: /custo
msearch/v1?key=dummy_google_api_key&cx=dummy_google_cx_key&q=How%20does%20GraphRAG%20work%3F&safe=active&start=1 (Caused by ProxyError('Unable to connect to pro
xy', OSError('Tunnel connection failed: 502 Tunnel Connection Failed')))

Wouldn't it be better to avoid this type of calls in the software, or set a parameter that avoids this possibility? I think it is essential to earn the trust of users.

[pengfeng]

Yes, if you want to do everything locally, just use the "local" retriever in all cases, which is more obviously from a UI (since retriever is required). On command line the default retriever is "google" and you need to specify -p retriever_type=local explicitly. Or you can specify an env variable EDS_WEB_RETRIEVER=local to set the default to local. We will make sure to add those to the documentation.

For the default search proxy, you can also set SEARCH_API_URL to your own search engines and corresponding API KEYs such as GOOGLE_API_KEY or BING_SEARCH_API_KEY. Because it is very difficult for people to set up the Google cloud account and get the keys, we provide a proxy as default to lower the adoption barrier. If the intention is pure local documents, then this part is not used at all.

Thanks for the great feedback! Local KB and private deployment are definitely key offering of this tool and we will make sure these points are clear in the documents.