OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
RAG & Knowledge Systems

Context Search Engine

Context Search Engine is a self-hosted Python application for indexing personal document folders and retrieving relevant passages through keyword, embedding-based or hybrid search. It returns source locations and evidence about how results ranked, rather than generating rewritten answers.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Context Search Engine — OpenViST project profile in RAG & Knowledge Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you want a self-hosted index for document discovery and need source passages and retrieval signals to inspect the results. It is particularly relevant for comparing keyword, dense and hybrid retrieval on your own corpus.

Who it is for

  • People who need to search their own PDF, Office, text, Markdown, HTML or CSV collections.
  • Researchers and teams who want to inspect matching passages and the retrieval evidence behind results.
  • Developers prototyping document retrieval or a search component for an internal RAG workflow.

Documented uses

  • Find passages in a folder of documents using terms, semantic similarity or both.
  • Inspect a result's source page, passage location, matched terms, retrievers and ranking signals.
  • Track additions, edits and deletions in monitored folders.
  • Compare retrieval strategies and evaluate them against labelled queries.
Before you choose

Requirements & limits

The repository identifies the project as open source under the MIT license. It is self-hosted and documents a Python 3.10+ setup.

  • Base installation supports keyword search; dense and hybrid search require the dense extra, including Torch and sentence-transformers.
  • The first dense search downloads the default BAAI/bge-small-en-v1.5 model; the README says later use works offline.
  • Scanned PDFs and images require the OCR extra.
  • Cross-encoder reranking is optional, off by default and described as more computationally expensive.
  • The application exposes ranking evidence and source locations rather than language-model-written summaries.

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

Set up the local search interface

Install the dense-search extra and start the documented UI and API service.

Before you start

  • Python 3.10 or later
  • A local copy of the repository
  • For dense search, the documented install includes Torch and sentence-transformers

Steps

  1. Clone the repository and change into its directory.

  2. Create and activate a Python virtual environment.

  3. Install the project with the dense extra using the command below.

    Documented command
    pip install -e ".[dense]"

  4. Run contextsearch doctor to check Python, SQLite FTS5, Torch and the model cache.

  5. Run contextsearch serve to start the UI and API at the documented local address.

Expected outcome

The local web UI and API are served at http://127.0.0.1:8080.

Guide limitations

  • The first dense search downloads the default embedding model, approximately 130 MB, to the Hugging Face cache.
  • On Linux, the default Torch installation may download a CUDA build several gigabytes in size; the README documents a CPU-build alternative.
  • Installing the base package without the dense extra provides keyword search only.
Documentation 1 ↗
Source-based practical guide

Index a folder and run a hybrid search

Index the bundled sample documents and search them using keyword and embedding retrieval.

Before you start

  • The repository installed with the dense extra
  • The contextsearch command available in the active environment

Steps

  1. Run contextsearch index examples/sample-docs to scan, chunk and embed the example files.

  2. Run contextsearch search "how do I get my money back" --strategy hybrid to retrieve matching content.

Expected outcome

The CLI prints an indexing summary and returns hybrid search results, with source passages and retrieval evidence described by the project.

Guide limitations

  • The first dense search downloads the default embedding model if it is not already cached.
  • The README describes these commands for bundled examples; it does not establish results for a particular external document collection.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
inboxpraveen/Context-Search-Engine · Official homepage title · retained in its source language
How we review sourcesSuggest a correction