OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Search & Research AI

Webglm

WebGLM is an open-source Python research implementation of web-enhanced question answering that retrieves web content, generates answers with a language model, and includes a human-preference-aware response scorer.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Webglm — OpenViST project profile in Search & Research AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use WebGLM when you want to study or experiment with question answering that combines web search and retrieval with language-model generation, and can prepare a local Python environment and search configuration.

Who it is for

  • Researchers studying web-enhanced question answering and human-preference-aware response scoring.
  • Developers exploring a self-hosted QA implementation with linked model checkpoints.
  • Teams conducting QA experimentation using the WebGLM-QA dataset.

Documented uses

  • Retrieve relevant web content to support answers to questions.
  • Generate responses using WebGLM model checkpoints.
  • Explore generator and retriever training or evaluate QA experiments.
  • Compare generated responses using a scorer designed to estimate quality in light of human preferences.
Before you choose

Requirements & limits

The repository is public and lists an Apache-2.0 license. The documented workflow is self-hosted; the README describes SerpAPI search using a key or Bing search through a local Playwright browser environment.

  • The system separates web retrieval, answer generation and preference-aware response scoring.
  • Search setup requires either a SerpAPI key or the documented Bing/Playwright route.
  • The README asks users to ensure sufficient device space but gives no hardware requirements.
  • The repository links 2B and 10B model checkpoints and a QA dataset.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Prepare WebGLM's local environment

Install the documented runtime components and prepare search access and the retriever checkpoint before trying the system.

Before you start

  • Python and the repository's requirements file
  • Node.js
  • Playwright and its browser dependencies
  • A SerpAPI key, or Bing search with a local Playwright browser environment
  • The WebGLM retriever checkpoint

Steps

  1. Clone the WebGLM repository and install its Python requirements with pip install -r requirements.txt.

  2. Install Node.js and the Playwright dependencies; the README gives apt install nodejs for Ubuntu and playwright install for Playwright.

  3. For SerpAPI search, obtain a key and set SERPAPI_KEY. Alternatively, use the documented Bing option with a local browser environment.

  4. Download the retriever checkpoint using python download.py retriever-pretrained-checkpoint, then set WEBGLM_RETRIEVER_CKPT to its path.

Expected outcome

The documented code, search configuration and retriever checkpoint are prepared for proceeding to the README's CLI or web-service instructions.

Guide limitations

  • The README says to ensure the device has enough space but gives no hardware or storage figures.
  • This covers preparation only; it does not claim that a run or deployment was completed.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
THUDM/WebGLM · Official homepage title · retained in its source language
How we review sourcesSuggest a correction