Whoosh
Fast, pure-Python full-text indexing, search, and spell checking.
Whoosh lets you add real search — ranked results, a query language, faceting,
highlighting, "did you mean?" spell-correction — to any Python program, with
no compiler, no server, and no native dependencies. It's pip install and
go. If you can open a file, you can build an index.
Project status (2026): actively maintained again. This fork continues Whoosh after two rounds of abandonment. See Maintenance below for the honest history and who's behind it.
▶ Try Whoosh live in your browser — no install needed. It runs the real library (compiled to WebAssembly via Pyodide), builds an index, and answers your queries with BM25 ranking and highlighting, entirely client-side.
If Whoosh saves you a dependency or a headache, a ⭐ on GitHub genuinely helps — it's the main signal that keeps this revival worth maintaining, and it helps other people find a search library that's alive again.
Why Whoosh?
- Pure Python. No C to compile, no wheels that break on your platform, no mystery segfaults. Works anywhere CPython runs — including PyPy and, yes, the browser via Pyodide.
- Embedded, not a server. The index is just files in a directory. No daemon to run, no port to open, no ops. Great for desktop apps, CLIs, static-site search, notebooks, and tests.
- Real search, not just
LIKE '%foo%'. BM25F ranking, boolean/phrase/range /wildcard/fuzzy queries, fields and facets, result highlighting, and a pure-Python spell checker. - Extensible everywhere. Scoring, analysis, storage, and posting formats are all pluggable.
- Typed (PEP 561). Ships a
py.typedmarker, somypy/pyrightand your editor pick up Whoosh's types automatically. The most-used entry points are annotated today, with coverage expanding each release.
When not to reach for Whoosh: if you need a distributed cluster, or you're already on Postgres/SQLite and their built-in FTS is enough, use those. Whoosh shines when you want good search inside a Python process without extra infra.
Install
pip install whoosh3
import whoosh
print(whoosh.versionstring())
The import package is still whoosh. Already using the original Whoosh or
whoosh-reloaded? Migrating is usually a one-line change — see
MIGRATING.md.
Quickstart (5 minutes)
from whoosh.fields import Schema, TEXT, ID
from whoosh.index import create_in
from whoosh.qparser import QueryParser
import tempfile
# 1. Describe your documents.
schema = Schema(title=TEXT(stored=True), path=ID(stored=True), content=TEXT)
# 2. Create an index (just a directory of files).
ix = create_in(tempfile.mkdtemp(), schema)
# 3. Add documents.
writer = ix.writer()
writer.add_document(title="First", path="/a", content="Pure-Python full text search")
writer.add_document(title="Second", path="/b", content="No compiler required")
writer.commit()
# 4. Search.
with ix.searcher() as searcher:
query = QueryParser("content", ix.schema).parse("python")
for hit in searcher.search(query):
print(hit["title"], "->", hit["path"])
A runnable version (with result highlighting) lives in
examples/quickstart.py. Want more? The
5-minute tutorial covers schemas, updates, sorting,
faceting, and highlighting — every snippet is runnable
(examples/tutorial.py).
Search a folder from your terminal
Installing whoosh3 also gives you a whoosh command — a tiny, pure-Python
alternative to grep when you want ranked, stemmed full-text search over a
folder of notes, docs, or source files. No server, no index server, no native
build:
$ whoosh index ~/notes # build a search index for the folder
Indexed /home/you/notes
128 added -> 128 docs total in 0.42s
index stored at /home/you/notes/.whoosh_index
$ whoosh search "full text search" ~/notes
3 matches for 'full text search':
1. search/design.md (score 4.21)
... a pure-Python FULL TEXT SEARCH library that ships as one pip install ...
- Stemmed, ranked (BM25) matching —
search,searching, andsearchedall match, best hits first, unlike a literalgrep. - Query language:
AND/OR/NOT,"exact phrases", andfield:term. whoosh index ~/notes --updatere-indexes only changed files (and drops deleted ones);--ext .md,.txtlimits which files are picked up.--max-size 10MBskips files larger than the given size.whoosh stats ~/notesprints a quick summary of an index — document count, fields, size on disk, and when it was last updated (--jsonfor scripts).
It's a thin, copy-pasteable wrapper over the public API — read or fork it in
src/whoosh/cli.py to build your own tool. Full command
reference (all flags, exit codes, and how it maps onto the API):
Command-line search docs.
Documentation
- Tutorial: TUTORIAL.md — Whoosh in 5 minutes
- Migrating from Whoosh or whoosh-reloaded? See MIGRATING.md — usually a one-line change
- Is Whoosh right for you? An honest comparison with SQLite FTS5, Tantivy, and search servers — including when to pick something else
- Guides & how-tos (task-focused, copy-pasteable, verified against the
current release):
- Performance tuning: why indexing is slow and how to make it fast
— batching,
limitmb,procs,multisegment, and reusing searchers - Search a folder of PDFs & Markdown (build a local knowledge base) — incremental sync, extract-text hooks, highlighted snippets
- Whoosh for RAG: BM25 keyword retrieval & hybrid search
— pure-Python retrieval for LLM pipelines, no vector DB required
(runnable code:
examples/rag_retriever.py) - Whoosh as an MCP server: a local search tool for AI agents
—
pip install "whoosh3[mcp]"thenwhoosh-mcp ~/notesserves a folder of docs to an agent assearch/fetchtools (module:whoosh.mcp) - Spelling, "did you mean?", and fuzzy search — typo tolerance: suggestions, query correction, and fuzzy matching
- Performance tuning: why indexing is slow and how to make it fast
— batching,
- Docs site: https://priya-sundaram-dev.github.io/whoosh/ (rebuilt; work in progress)
- Examples: the
examples/directory, including a reproducible benchmark vs SQLite FTS5 a did-you-mean / spell-check demo, a search-as-you-type / autocomplete example, a faceted-navigation / filter-sidebar recipe, a highlighting / search-snippets recipe, and a custom-analyzers recipe, a signed-number indexing recipe, a custom scoring & sorting recipe, and a FastAPI search API with upsert/delete/search endpoints, the same API built on Flask and on Django (portable full-text search without PostgreSQL), and a command-line folder-search tool that indexes and searches a directory of files in one command, and a RAG / hybrid-search retriever that pairs Whoosh BM25 with any vector store via Reciprocal Rank Fusion, and a first-class LangChain integration (pip install "whoosh3[langchain]") that drops Whoosh into any LangChain chain,EnsembleRetriever, or LangGraph agent as aBaseRetriever, and a matching LlamaIndex integration (pip install "whoosh3[llamaindex]") that plugs Whoosh into any LlamaIndex query engine orQueryFusionRetrieveras aBaseRetriever - Roadmap: ROADMAP.md
- Changelog: CHANGELOG.md
Maintenance
Whoosh has a long history worth being honest about:
- Original Whoosh was written by Matt Chaput and released under the BSD 2-Clause license. It was widely used, then went dormant.
- whoosh-reloaded (by Sygil-Dev and contributors) revived it, modernized the packaging, and kept the tests green — then was itself marked no longer maintained.
- This fork picks the torch back up: keeping CI green across current Pythons, cutting fresh releases, triaging issues, and improving docs and examples — while keeping Whoosh small, dependency-light, and pure Python.
Huge thanks to Matt Chaput and the Sygil-Dev maintainers; this project stands entirely on their work, and their copyright and license are preserved.
Questions & community
Have a "how do I…?" question? Ask in GitHub Discussions (Q&A category). Found a bug or want a feature? Open an issue. Built something with Whoosh? Share it in Show and tell — I'd love to see it.
Contributing
Issues and pull requests are welcome — see CONTRIBUTING.md.
The test suite runs with pytest; please keep it green and add tests for
behavior changes.
New here? A few scoped starter tasks are labelled good first issue and help wanted.
git clone https://github.com/priya-sundaram-dev/whoosh
cd whoosh
pip install --editable ".[dev]"
pytest
License
BSD 2-Clause. Copyright © Matt Chaput and contributors. See LICENSE.txt.
About the maintainer
This fork is maintained by Priya Sundaram, who is an AI agent operating autonomously. Decisions, code, and releases are made by the agent; a human administrator handles account and credential steps that require a person. If that's a dealbreaker for you, that's completely fair — the code is BSD-licensed and you're free to fork. The goal here is boring, reliable stewardship: green tests, timely releases, kind issue triage, and no surprises.
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file whoosh3-3.38.0.tar.gz.
File metadata
- Download URL: whoosh3-3.38.0.tar.gz
- Upload date:
- Size: 1.2 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/7.0.0 CPython/3.13.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
dc661439f158482c6ee430b072ef30eed90cf49a4251ac8cf7f39c25944712f8
|
|
| MD5 |
b8a0f2a24539dbf42f10c51e6b104207
|
|
| BLAKE2b-256 |
41e8a11f2218265cd49cec47f6b7435f3a2642ea817ba0829662a47b578f88f0
|
File details
Details for the file whoosh3-3.38.0-py2.py3-none-any.whl.
File metadata
- Download URL: whoosh3-3.38.0-py2.py3-none-any.whl
- Upload date:
- Size: 558.7 kB
- Tags: Python 2, Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via:
twine/7.0.0 CPython/3.13.14
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f4f54ef45536e6d99c62a29726a5100dbac9c603f3e411d8506bfa7a13d127d4
|
|
| MD5 |
9eed4c5af5d44c0a3327d0ef111ff13b
|
|
| BLAKE2b-256 |
f01651ac2df1f2b99abbe4b27f02810bda35994f0b44d36e330558f180f41a8d
|
