Repolode
Submit
mamei16

mamei16/LLM_Web_search

An extension for text-generation-webui that lets a local LLM search the web and append relevant page excerpts to its answers.

View on GitHub
Stars
284
Forks
40
Open issues
6
Last commit
3 days ago

What is it?

LLM_Web_search gives local language models the ability to search the web inside oobabooga's text-generation-webui. The model can trigger a search either through native tool calling or through a regular expression that matches a command in its output. The extension fetches the result pages, then uses a dense embedding model combined with Okapi BM25 (or optionally SPLADE) to pull out the passages relevant to the query before appending them to the model's output.

Search can run through DuckDuckGo by default or through a SearXNG instance that returns JSON. It offers a simple mode that uses only the snippets from search results, and a full mode that scans entire pages. Page text can be split into chunks using fixed-size splitting, semantic splitting based on embedding distance, or a DistilBERT-based token classification model.

The project targets users who run models locally with text-generation-webui and want them to answer questions about current information. It is installed as a standard extension and is licensed under AGPL-3.0.

Key features

  • 01Native tool calling or regex-triggered search commands
  • 02DuckDuckGo by default, with optional SearXNG backend
  • 03Simple and full search modes
  • 04Okapi BM25 or SPLADE keyword retrieval combined with dense embeddings
  • 05Character, semantic, and token-classification chunking options

Use cases

  • Answering questions about recent events with a locally hosted model
  • Giving a local chat model web access without sending prompts to a hosted service
  • Running private search through a self-hosted SearXNG instance

Quick start

  1. 1

    Install the extension from the Session tab of the web UI

  2. 2

    Run the update_wizard script and choose Install/update extensions requirements

  3. 3

    Start the server with the extension enabled

    python server.py --extension LLM_Web_search

Things to know

  • It is an extension for text-generation-webui, so that application must be installed first.
  • Licensed under AGPL-3.0, a copyleft license that requires derivative network services to share source code.
  • Full search, semantic chunking and SPLADE are slower and need more resources, with SPLADE benefiting from a GPU.

Topics

#llm#web-search#rag#text-generation-webui#duckduckgo#searxng#bm25#embeddings#chatgpt#faiss

FAQ

What is LLM_Web_search?
An extension for text-generation-webui that lets a local LLM search the web and append relevant page excerpts to its answers. LLM_Web_search gives local language models the ability to search the web inside oobabooga's text-generation-webui. The model can trigger a search either through native tool calling or through a regular expression that matches a command in its output. The extension fetches the result pages, then uses a dense embedding model combined with Okapi BM25 (or optionally SPLADE) to pull out the passages relevant to the query before appending them to the model's output.
Is LLM_Web_search free to use? What license does it have?
LLM_Web_search is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.
Is LLM_Web_search still maintained?
Actively maintained. The last commit to LLM_Web_search was on Nov 3, 2026.
What can I use LLM_Web_search for?
Answering questions about recent events with a locally hosted model Giving a local chat model web access without sending prompts to a hosted service Running private search through a self-hosted SearXNG instance
How do I install LLM_Web_search?
Install the extension from the Session tab of the web UI Run the update_wizard script and choose Install/update extensions requirements Start the server with the extension enabled: python server.py --extension LLM_Web_search
Where can I get LLM_Web_search?
The source code and installation instructions are on GitHub: https://github.com/mamei16/LLM_Web_search

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects