OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Research Agents

Searchlm

SearchLM is a public Python research repository containing code for ExSearch, an approach that uses large language models as agentic searchers. The system iteratively seeks information, selects useful knowledge, records evidence and summarizes an answer, with repository code for evaluation and training.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Searchlm — OpenViST project profile in Research Agents.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use SearchLM when you want code and research workflows for experimenting with LLMs as iterative search agents, particularly where the goal is to examine reasoning, retrieval and evidence collection together.

Who it is for

  • Researchers exploring LLM reasoning during search and retrieval-augmented question answering.
  • Developers prototyping an LLM-driven search agent or adapting the repository's training and evaluation workflows.

Documented uses

  • Study an iterative search process in which an LLM gathers information and retains useful evidence.
  • Prepare a retrieval module using a Wikipedia corpus and ColBERT or a customized retriever.
  • Run the documented warm-up, trajectory exploration and re-weighted trajectory-learning stages with model and dataset inputs.
Before you choose

Requirements & limits

The repository is public and specifies an Apache-2.0 license.

  • ExSearch is a research codebase with training and evaluation scripts, not a hosted search service.
  • The documented retrieval setup uses a Wikipedia corpus and ColBERT, while allowing a locally deployed customized retrieval model.
  • Training examples use CUDA devices and distributed training; model-library versions vary with the selected LLM backbone.
  • Weights & Biases login is optional when using it to record training loss.

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 the documented ExSearch environment and retrieval module

Set up the Python environment and retrieval resources described in the repository before following its training or evaluation workflows.

Before you start

  • Conda and Python 3.10
  • The repository's requirements.txt
  • A Wikipedia document corpus and a ColBERT retrieval model/checkpoint
  • A chosen LLM backbone; compatible PyTorch, vLLM and Transformers versions depend on that model

Steps

  1. Create and activate a Conda environment with Python 3.10, then install the packages in requirements.txt.

  2. Obtain the Wikipedia corpus using the linked DPR data resource.

  3. Set up ColBERT or another customized retrieval model locally to preprocess the dataset, following the retrieval folder's README.

  4. Select model-library versions for the backbone; the README suggests vLLM 0.6.3, torch 2.4.0+cu118 and transformers 4.45.0 for Qwen and Mistral.

Expected outcome

The documented Python environment and retrieval resources are prepared for the repository's subsequent workflows.

Guide limitations

  • This guide covers setup only; it does not establish that a particular machine or configuration will run successfully.
  • The README allows model-library versions to vary by backbone.
  • The training examples use CUDA devices and distributed-training commands.
Documentation 1 ↗
Evidence and transparency

Sources & review

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